Plaintiffs' Corrected Omnibus Opposition to Defendants' Motions for Summary Judgment, In Re: Social Media Adolescent Addiction/Personal Injury Products Liability Litigation

Plaintiffs' filing21 Nov 2025

Transcribed text as recorded, including headers and formatting fragments.

Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 1 of 235 1 [Parties and Counsel Listed on Signature Pages] 2 3 4 5 6 7 8 UNITED STATES DISTRICT COURT 9 FOR THE NORTHERN DISTRICT OF CALIFORNIA 10 IN RE: SOCIAL MEDIA ADOLESCENT MDL No. 3047 11 ADDICTION/PERSONAL INJURY PRODUCTS LIABILITY LITIGATION Case No. 4:22-md-03047-YGR (PHK) 12 This Document Relates To: PLAINTIFFS' CORRECTED OMNIBUS 13 OPPOSITION TO DEFENDANTS' Breathitt County Board of Education v. Meta MOTIONS FOR SUMMARY 14 Platforms, Inc., et al. JUDGMENT 15 Tucson Unified School District v. Meta Platforms, Judge: Hon. Yvonne Gonzalez Rogers 16 Inc., et al. Magistrate Judge: Hon. Peter H. Kang 17 Charleston County School District v. Meta Date: January 26, 2026 Platforms, Inc., et al. Time: 8:00 AM 18 Place: Courtroom 1, 4th Floor Irvington Public Schools v. Meta Platforms, Inc., et 19 al. 20 Dekalb County School District v. Meta Platforms, 21 Inc., et al. 22 Board of Education of Harford County v. Meta 23 Platforms, Inc., et al. 24 25 26 27 28 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 2 of 235 1 TABLE OF CONTENTS 2 I. INTRODUCTION AND "ROAD MAP" ........................................................................... 1 3 II. LEGAL STANDARD ......................................................................................................... 3 4 III. ARGUMENT ...................................................................................................................... 3 5 A. Genuine material disputes of fact concerning Defendants' liability preclude summary judgment.................................................................................................. 3 6 1. Defendants are liable to the Districts for failing to warn about the risks of 7 their platforms. ............................................................................................ 3 8 a) The Court's prior rulings establish Defendants owed a duty to the Districts. .......................................................................................... 4 9 b) Defendants owe a duty to warn the Districts directly. .................... 5 10 c) Defendants owe a duty to warn students and parents about their 11 platforms' dangers. ......................................................................... 8 12 d) Defendants' failure to warn caused the Districts' injuries. ........... 10 13 2. Defendants breached their duty of care. ................................................... 12 14 a) Meta's Conduct ............................................................................. 12 15 (1) Meta's business model ...................................................... 12 16 (2) Meta's targeting of school-aged children ......................... 14 17 (3) Meta's targeting of schools ............................................... 17 18 (4) Meta's knowledge of harms .............................................. 21 19 (5) Meta's failure to exercise reasonable care ........................ 33 20 (6) Meta's misrepresentations and manipulations .................. 68 21 (7) Meta's willful and knowing disregard of the rights of others ........................................................................................... 74 22 b) TikTok's Conduct ......................................................................... 77 23 (1) TikTok's business model .................................................. 77 24 (2) TikTok's targeting of school-aged children ...................... 79 25 (3) TikTok's targeting of schools ........................................... 82 26 (4) TikTok's knowledge of harms .......................................... 86 27 (5) TikTok's failure to exercise reasonable care .................... 93 28 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 3 of 235 1 (6) TikTok's misrepresentations and manipulations ............ 114 2 (7) TikTok failed to warn parents, educators, and schools about known mental health risks .............................................. 115 3 c) Snap's Conduct ........................................................................... 116 4 (1) Snapchat's business model ............................................. 116 5 (2) Snapchat's targeting of school-aged children ................. 118 6 (3) Snapchat's targeting of schools ...................................... 121 7 (4) Snap's knowledge and willful blindness of harms ......... 124 8 (5) Snap's failure to exercise reasonable care ...................... 130 9 (6) Snap's misrepresentations and manipulations ................ 146 10 (7) Snap failed to warn parents, educators, and schools about 11 known mental health risks .............................................. 148 12 d) Google/YouTube's Conduct ....................................................... 150 13 (1) YouTube's business model ............................................. 150 14 (2) YouTube's targeting of school-aged children................. 151 15 (3) YouTube's targeting of schools ...................................... 155 16 (4) YouTube's knowledge of harms ..................................... 162 17 (5) YouTube's failure to exercise reasonable care ............... 165 18 (6) YouTube's misrepresentations and manipulations ......... 184 19 (7) YouTube's failure to warn .............................................. 185 20 3. Defendants' conduct substantially interferes with the public right to health, safety, and education in Arizona, Georgia, Kentucky, and Maryland .... 185 21 B. Defendants' causation arguments are premised on disputed facts and legal errors. 22 ............................................................................................................................. 187 23 1. Defendants improperly raise the evidentiary bar in requiring Plaintiffs to quantify specific student usage ............................................................... 187 24 2. Defendants' "entanglement" arguments on causation are premised on legal 25 errors ....................................................................................................... 190 26 a) Defendants misapprehend the features at issue by writing this Court's failure to warn ruling out of existence. .......................... 190 27 b) The Districts need not rule out every possible alternative causal 28 factor to avoid Section 230. ........................................................ 191 ii PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 4 of 235 1 c) Juries are routinely called upon to evaluate liability in the context of concurrent causation ............................................................... 193 2 C. Genuine material disputes of fact concerning the Districts' damages preclude 3 summary judgment.............................................................................................. 195 4 1. Each Plaintiff has presented evidence of compensable "hard costs." ..... 195 5 a) The Districts presented evidence of mitigation measures, including mental health resources, and technology materials, systems, and 6 programming ............................................................................... 196 7 b) Defendants cross-cutting entanglement arguments do not support summary judgment on causation ................................................ 199 8 2. Defendants can be required to compensate Districts for the diversion of 9 school resources caused by their misconduct. ........................................ 201 10 3. Mr. Klein's survey raises disputed issues of fact regarding the Districts' lost time damages. ................................................................................... 204 11 D. The Districts' claims for future damages are legally cognizable. ....................... 205 12 1. The Districts proffer relevant evidence supporting a claim for future 13 damages................................................................................................... 205 14 2. Breathitt, DeKalb, Harford, and Tucson's nuisance claims permit the abatement they seek. ............................................................................... 210 15 a) Abatement is a distinct equitable remedy designed to eliminate 16 ongoing and future public harms. ............................................... 210 17 b) Dr. Hoover's strategic plans provide the framework for crafting an abatement remedy. ...................................................................... 211 18 (1) The abatement remedy may include monetary relief. ..... 212 19 (2) Defendants' state law authorities are inapposite. ............ 214 20 (3) Defendants' attacks on Dr. Hoover's plans lack merit. .. 215 21 3. The Districts' harms are not derivative. .................................................. 216 22 IV. CONCLUSION ............................................................................................................... 217 23 24 25 26 27 28 iii PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 5 of 235 1 TABLE OF AUTHORITIES 2 Cases 3 Aaron's, Inc. v. MDC Grp., Inc., 2010 WL 11505919 (N.D. Ga. Dec. 15, 2010). ................................. 202 4 Accessory Overhaul Grp., Inc. v. Mesa Airlines, Inc., 994 F. Supp. 2d 1296 (N.D. Ga. 2014) ............. 207 5 Adams v. Comm'rs of Town of Trappe, 102 A.2d 830 (Md. Ct. App. 1954).......................................... 214 6 Alcantar v. Hobart Serv., 2013 WL 156530 (C.D. Cal. Jan. 15, 2013) .................................................. 205 7 Allen v. Greenville Hotel Partners, Inc., 2006 WL 1817804 (D.S.C. June 30, 2006) ........................... 194 8 Allen v. Long Mfg. NC, Inc., 505 S.E.2d 354 (S.C. Ct. App. 1998) ......................................................... 11 9 Am. Motorcycle Ass'n v. Super. Ct., 578 P.2d 899 (Cal. 1978) .............................................................. 193 10 Anderson v. Liberty Lobby, Inc., 477 U.S. 242 (1986) ............................................................................... 3 11 Andino v. Apple, Inc., 2021 WL 1549667 (E.D. Cal. Apr. 20, 2021) ................................................ 210-11 12 Ass'n of Washington Pub. Hosp. Dists. v. Philip Morris Inc., 241 F.3d 696 (9th Cir. 2001) ................ 216 13 Atari Interactive, Inc. v. Redbubble, Inc., 515 F. Supp. 3d 1089 (N.D. Cal. 2021) ................................... 3 14 Bailey v. N. Am. Refractories Co., 95 S.W.3d 868 (Ky. Ct. App. 2001) ................................................ 190 15 Ballou v. Sigma Nu Gen. Fraternity, 352 S.E.2d 488 (S.C. Ct. App. 1986) .......................................... 194 16 Barnes v. Yahoo!, Inc., 570 F.3d 1096, 1107 (9th Cir. 2009) ................................................................... 69 17 Barrett Props., LLC v. Roberts Capitol, Inc., 729 S.E.2d 621 (Ga. Ct. App. 2012) .............................. 194 18 Battaglia v. United Parcel Serv., Inc., 70 A.3d 602 (N.J. 2013) ............................................................ 206 19 Beasley v. Astrue, 2011 WL 1327130 (W.D. Wash. Mar. 24, 2011)...................................................... 186 20 Berman v. Freedom Fin. Network, LLC, 400 F. Supp. 3d 964 (N.D. Cal. 2019) ................................... 195 21 Blondell v. Courtney Station 300 LLC, 865 S.E.2d 589 (Ga. Ct. App. 2021) ........................................ 187 22 Brandes v. Mitterling, 196 P.2d 464 (Ariz. 1948) .................................................................................. 214 23 Bride v. Yolo Technologies, Inc., 112 F.4th 1168 (9th Cir. 2024)……………………………………...192 24 Brookside Assocs. v. Rifkin, 49 F.3d 490 (9th Cir. 1995) ........................................................................... 3 25 Brown v. City of Phoenix, 557 P.3d 321 (Ariz. Ct. App. 2024).............................................................. 214 26 Bullock v. Volkswagen Grp. of Am., Inc., 107 F. Supp. 3d 1305 (M.D. Ga. 2015) .................................. 11 27 Cactus Corp. v. State ex rel. Murphy, 480 P.2d 375 (Ariz. Ct. App. 1971) ........................................... 214 28 iv PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 6 of 235 1 Calise v. Meta Platforms, Inc., 103 F.4th 732 (9th Cir. 2024) ............................................................... 192 2 Campione v. Soden, 695 A.2d 1364 (N.J. 1997)..................................................................................... 200 3 Carter v. Wallace & Gale Asbestos Settlement Tr., 96 A.3d 147 (Md. Ct. App. 2014) ................. 200, 207 4 Celotex Corp. v. Catrett, 477 U.S. 317 (1986) ........................................................................................... 3 5 Cepelak v. HP Inc., 2021 WL 5298022 (N.D. Cal. Nov. 15, 2021) ....................................................... 211 6 CertainTeed Corp. v. Fletcher, 794 S.E.2d 641 (Ga. 2016) ....................................................................... 7 7 Charleston Lumber Co. v. Miller Hous. Corp., 458 S.E.2d 431 (S.C. Ct. App. 1995) .......................... 202 8 City & Cnty. of San Francisco v. Purdue Pharma L.P., 620 F. Supp. 3d 936, 1003 (N.D. Cal. 2022) ....... 9 .............................................................................................................................................. 187-188, 189 10 City of Huntington, W. Virginia v. AmerisourceBergen Drug Corp., 2025 WL 3009526 (4th Cir. Oct. 28, 11 2025) ........................................................................................................................................... 210, 212 12 City of Safford v. Seale, 2009 WL 3390172 (Ariz. Ct. App. Oct. 21, 2009) .......................................... 214 13 Coleman v. Mondelez Int'l Inc., 554 F. Supp. 3d 1055 (C.D. Cal. 2021)....................................... 210, 211 14 Colon v. Robinson, 2014 WL 7466566 (N.J. Super. Ct. App. Div. Jan. 6, 2015)……………………...206 15 Comdyne I, Inc. v. Corbin, 908 F.2d 1142 (3d Cir. 1990) ...................................................................... 202 16 Convoy Co. v. Sperry Rand Corp., 672 F.2d 781 (9th Cir. 1982) .................................................. 201, 204 17 Correct Transmission, LLC v. Nokia of Am. Corp., 2024 WL 1289821 (E.D. Tex. Mar. 26, 2024) ..... 201 18 Coury Bros. Ranches, Inc. v. Ellsworth, 446 P.2d 458 (Ariz. 1968) ...................................................... 207 19 CRS Sirrine, Inc. v. Dravo Corp., 464 S.E.2d 897 (Ga. Ct. App. 1995) ................................................ 194 20 Davidson v. Miller, 344 A.2d 422 (Md. 1975) ....................................................................................... 206 21 DiLeo v. Nugent, 592 A.2d 1126 (Md. Ct. Spec. App. 1991)…………………………………………..206 22 Doe (K.B.) v. Backpage.com, LLC, 724 F. Supp. 3d 882 (N.D. Cal. 2024) ........................................... 193 23 Doe 1 v. Mayorkas, 530 F. Supp. 3d 893 (N.D. Cal. 2021) ........................................................................ 5 24 Doe 1 v. Twitter, 148 F.4th 635 (9th Cir. 2025) ..................................................................................... 192 25 Doe v. Grindr Inc., 128 F.4th 1148 (9th Cir. 2025) ................................................................................ 193 26 Doe v. Internet Brands, Inc., 824 F.3d 846 (9th Cir. 2016) ...................................................................... 69 27 Dozier Crane & Mach., Inc. v. Gibson, 644 S.E.2d 333 (Ga. Ct. App. 2007) ........................................... 6 28 Dunn Appraisal Co. v. Honeywell Info. Sys. Inc., 687 F.2d 877 (6th Cir. 1982) ........................... 202, 204 v PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 7 of 235 1 Fed. Deposit Ins. Corp. v. Loudermilk, 826 S.E.2d 116 (Ga. 2019)....................................................... 200 2 Fortune Dynamic, Inc. v. Victoria's Secret Stores Brand Mgmt., Inc., 618 F.3d 1025 (9th Cir. 2010) . 205 3 Fuller v. Idaho Dep't of Corr., 865 F.3d 1154 (9th Cir. 2017) ............................................................... 209 4 Furlong v. Dyal, 539 S.E.2d 836 (Ga. Ct. App. 2000) ........................................................................... 189 5 Gassaway Constr. Co. v. Gentry, 264 S.W.2d 658 (Ky. 1954) .............................................................. 202 6 Georgia Cas. & Sur. Co. v. Salter's Indus. Serv., Inc., 734 S.E.2d 415 (Ga. Ct. App. 2012) .................. 11 7 Gilbert v. Stewart, 255 A.3d 1101 (N.J. 2021) ............................................................... 187, 190, 194, 196 8 Gilmore v. Cohen, 386 P.2d 81 (Ariz. 1963) .......................................................................................... 207 9 Gipson v. Kasey, 150 P.3d 228 (2007)........................................................................................................ 4 10 Gourdine v. Crews, 955 A.2d 769 (Md. 2008) ........................................................................................... 7 11 Green Meadows Hous. Partners, LP v. Macon-Bibb Cnty., 906 S.E.2d 430 (2024) ............................. 215 12 Griffey v. Magellan Health Inc., 562 F. Supp. 3d 34 (D. Ariz. 2021) .................................................... 203 13 Guthrie v. Transamerica Life Ins. Co., 561 F. Supp. 3d 869 (N.D. Cal. 2021)...................................... 214 14 Haltiwanger v. Barr, 186 S.E.2d 819 (S.C. 1972) .......................................................................... 205, 206 15 Hanna v. Ward Manuf., Inc., 2016 WL 3196467 (M.D. Fla. June 9, 2016) ............................................ 11 16 Hayes v. Crawford, 730 S.E.2d 26 (Ga. Ct. App. 2012)......................................................................... 190 17 Hickerson v. Yamaha Motor Corp., U.S.A., 2016 WL 4367141 (D.S.C. Aug. 16, 2016) ........................ 11 18 Hirsh v. Manley, 300 P.2d 588 (Ariz. 1956) ........................................................................................... 207 19 Hunt v. City of L.A., 638 F.3d 703 (9th Cir. 2011) ................................................................................. 189 20 IceMOS Tech. Corp. v. Omron Corp., 2020 WL 1083817 (D. Ariz. Mar. 6, 2020)............................... 203 21 In re Blackbaud, Inc., Customer Data Breach Litig., 567 F. Supp. 3d 667 (D.S.C. 2021) .................... 203 22 In re Gen. Motors LLC Ignition Switch Litig., 339 F. Supp. 3d 262 (S.D.N.Y. 2018) ........................... 203 23 In re JUUL Labs, Inc., Mktg., Sales Pracs., & Prods. Liab. Litig., 497 F. Supp. 3d 552 (N.D. Cal. 2020) 24 ......................................................................................................................................................... 8, 212 25 In re Nat'l Prescription Opiate Litig., 2019 WL 4043938 (N.D. Ohio Aug. 26, 2019) ............ 210, 212-13 26 In re Nat'l Prescription Opiate Litig., 622 F. Supp. 3d 584 (N.D. Ohio 2022), rev'd on other grounds, 27 2025 WL 354758 (6th Cir. Jan. 31, 2025) .......................................................................................... 213 28 In re Nat'l Prescrip. Opiate Litig., 589 F. Supp. 3d 790 (N.D. Ohio 2022)………………………186, 189 vi PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 8 of 235 1 In re Nat'l Prescription Opiate Litig., 2019 WL 4178617 (N.D. Ohio Sept. 3, 2019) ................... 188, 190 2 In re Soc. Media Adolescent Addiction/Pers. Inj. Prods. Liab. Litig., 754 F. Supp. 3d 946 3 (N.D. Cal. 2024)............................................................................................................................. passim 4 In re Soc. Media Adolescent Addiction/Pers. Inj. Prods. Liab. Litig., 2024 WL 5694360 5 (N.D. Cal. Nov. 15, 2024) ................................................................................................................... 185 6 In re Soc. Media Adolescent Addiction/Pers. Inj. Prods. Liab. Litig., 753 F. Supp. 3d 849 7 (N.D. Cal. 2024).................................................................................................................................. 190 8 In re Williams Sec. Litig.-WCG Subclass, 558 F.3d 1130 (10th Cir. 2009)………….….……………..193 9 J&W Corp. v. Broad Creek Marina, LLC, 896 S.E.2d 328 (S.C. Ct. App. 2023) .................................. 207 10 J.D. Jewell, Inc. v. Hancock, 175 S.E.2d 847 (Ga. 1970)....................................................................... 214 11 J.N. Legacy Grp., Inc. v. City of Dallas, 745 S.E.2d 721 (Ga. Ct. App. 2013) ...................................... 206 12 J.T. Baggerly v. CSX Transp., Inc., 635 S.E.2d 97 (S.C. 2006) ..................................................... 189, 190 13 Johnson v. Wood, 2025 WL 2715888 (Ga. Ct. App. Sept. 24, 2025) ..................................... 187, 195, 196 14 Jones by & through Jones v. IC Bus, LLC, 626 S.W.3d 661 (Ky. Ct. App. 2020) ................................... 10 15 Jones v. Hutchinson Mfg., Inc., 502 S.W.2d 66 (Ky. 1973) ....................................................................... 7 16 KeraLink Int'l, Inc. v. Geri-Care Pharms. Corp., 60 F.4th 175 (4th Cir. 2023) .................................... 203 17 KeraLink Int'l, Inc. v. Stradis Healthcare, LLC, 2021 WL 5832242 (D. Md. Nov. 2, 2021) ................ 203 18 Kennedy Krieger Inst. v. Partlow, 191 A.3d 425, (Md. 2018)……………………………………………8 19 Kuciemba v. Victory Woodworks, 531 P. 3d 924 (Cal. 2023) .................................................................... 7 20 Lemmon v. Snap, Inc., 995 F.3d 1085 (9th Cir. 2021)……………….………………………...… 191, 192 21 Lurry v. PharMerica Corp., 2024 WL 2965642 (W.D. Ky. June 12, 2024) .......................................... 202 22 MacDonald v. Patriot, LLC, 2017 WL 1788115 (Md. Ct. Spec. App. May 5, 2017) ............................ 203 23 Martin v. Howard Cnty., 709 A.2d 125 (Md. 1998) ............................................................................... 214 24 Mason v. Arizona Pub. Serv. Co., 622 P.2d 493 (Ariz. Ct. App. 1980) ................................................. 189 25 Masson v. New Yorker Mag., Inc., 501 U.S. 496 (1991) ............................................................................ 3 26 Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574 (1986).................................................. 3 27 Mauro v. Raymark Indus., Inc., 561 A.2d 257 (N.J. 1989) .................................................................... 206 28 May v. Holzknecht, 320 S.W.3d 123 (Ky. Ct. App. 2010) ..................................................................... 205 vii PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 9 of 235 1 Mayer v. Hous. Auth. of Jersey City, 202 A.2d 439 (N.J. Super. Ct. App. Div. 1964) .......................... 194 2 Mayor and City of Baltimore v. Purdue Pharma, L.P., No. 24-C-18-000515 (Md. Cir. Ct. Aug. 8, 2025) 3 ............................................................................................................................................................. 213 4 McAlister v. Loeb & Loeb, LLP, 571 P.3d 891 (Ariz. 2025) .................................................................. 207 5 McNutt Oil & Refin. Co. v. D'Ascoli, 281 P.2d 966 (Ariz. 1955) ......................................................... 207 6 Meraz v. Ford Motor Co., 2014 WL 12558123 (C.D. Cal. June 13, 2014) ........................................... 187 7 Mitsubishi Int'l Corp. v. Cardinal Textile Sales, Inc., 14 F.3d 1507 (11th Cir. 1994) ........................... 211 8 Mobile Conversions, Inc. v. Allegheny Ford Truck Sales, 2014 WL 7369898 (W.D. Pa. Dec. 29, 2014) 9 ............................................................................................................................................................. 202 10 Moody v. NetChoice, LLC, 603 U.S. 707 (2024) ...................................................................................... 36 11 Moore v. City of Atlanta, Ga., 2022 WL 19517299 (N.D. Ga. Dec. 19, 2022) ...................................... 200 12 Musacchio v. United States, 577 U.S. 237 (2016) ...................................................................................... 5 13 Nelson v. Am. Honda Motor Co., 2021 WL 2877919 (W.D. Pa. May 17, 2021) ..................................... 11 14 NetChoice, LLC v. Bonta, 2025 WL 2600007 (9th Cir. Sept. 9, 2025) .................................................... 36 15 New Jersey Dep't of Env't Prot. v. E.I. du Pont de Nemours & Co., 2021 WL 6144081 (D.N.J. Dec. 30, 16 2021) ....................................................................................................................................................... 7 17 New York v. Ferber, 458 U.S. 747 (1982) ................................................................................................ 69 18 Nobles v. Jiffy Mkt. Food Store Corp., 579 S.E.2d 63 (Ga. Ct. App. 2003) ........................................... 207 19 Nolan v. Starlight Pines Homeowners Ass'n, 216 Ariz. 482 (Ct. App. 2007) ................................... 186-87 20 Nuveen Mun. High Income Opportunity Fund v. City of Alameda, 730 F.3d 1111 (9th Cir. 2013)....... 193 21 Ogden v. Bumble Bee Foods, LLC, 2014 WL 27527 (N.D. Cal. Jan. 2, 2014) .......................................... 3 22 Ostrovskaya v. St. John Knits, Inc., 2022 WL 2102895 (C.D. Cal. Mar. 31, 2022) ............................... 211 23 Owens Corning Fiberglas Corp. v. Parrish, 58 S.W.3d 467 (Ky. 2001) ............................................... 200 24 Parris v. 3M Co., 595 F. Supp. 3d 1288 (N.D. Ga. 2022) .......................................................................... 9 25 Pathways, Inc. v. Hammons, 113 S.W.3d 85 (Ky. 2003) ............................................................... 187, 196 26 People v. ConAgra Grocery Prods. Co., 227 Cal. Rptr. 3d 499 (Ct. App. 2017) .................................. 214 27 Pierce v. Johns-Manville Sales Corp., 464 A.2d 1020 (1983) ............................................................... 206 28 Pinckney v. Meta Platforms Inc., 127 F.4th 516 (4th Cir. 2025) ............................................................ 193 viii PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 10 of 235 1 Piner v. Sup. Ct. in and for Cnty. of Maricopa, 962 P.2d 909 (Ariz. 1998) ........................................... 200 2 Pittway Corp. v. Collins, 973 A.2d 771 (Md. Ct. App. 2009) ........................................................ 187, 196 3 PivotHealth Holdings LLC v. Horton, 2025 WL 1865788 (D. Ariz. July 7, 2025)................................ 202 4 Pratt v. Amisub of SC, Inc., 912 S.E.2d 268 (S.C. Ct. App. 2025) ......................................................... 200 5 Quinalty v. FocusIT LLC, 2024 WL 342454 (D. Ariz. Jan. 30, 2024) ................................................... 203 6 Quiroz v. ALCOA Inc., 416 P.3d 824 (Ariz. 2018) ..................................................................................... 7 7 Quisenberry v. Huntington Ingalls Inc., 818 S.E.2d 805 (Va. 2018) ......................................................... 5 8 Quynn v. Hulsey, 850 S.E.2d 725 (Ga. 2020) ......................................................................................... 200 9 R & R Insulation Servs. v. Royal Indem. Co., 705 S.E. 2d 223 (Ga. Ct. App. 2010) ............................... 10 10 Ramah Navajo Sch. Bd., Inc. v. Sebelius, 2013 WL 12303945 (D.N.M. May 9, 2013) ......................... 194 11 Rancho Pescado, Inc. v. Nw. Mut. Life Ins. Co., 680 P.2d 1235 (Ariz. Ct. App. 1984) ......................... 207 12 Rhynes v. Stryker Corp., 2011 WL 2149095 (N.D. Cal. May 31, 2011) ................................................ 211 13 Robertson v. Sixpence Inns of Am., Inc., 789 P.2d 1040 (Ariz. 1990) .................................................... 194 14 Robinson v. C.R. Bard, Inc., 2016 WL 3361825 (N.D. Cal. June 17, 2016) .......................................... 211 15 Roper v. Big Heart Pet Brands, Inc., 510 F. Supp. 3d 903 (E.D. Cal. 2020) ......................................... 211 16 Rowhouses, Inc. v. Smith, 133 A.3d 1054 (Md. 2016) ............................................................................ 189 17 Ruggiero v. Yamaha Motor Corp., U.S.A., 2017 WL 1197755 (D.N.J. Mar. 31, 2017) .......................... 11 18 Saide v. Stanton, 659 P.2d 35 (Ariz. 1983) ............................................................................................. 206 19 Schall v. Suzuki Motor of Am., Inc., 450 F. Supp. 3d 771 (W.D. Ky. 2020) .......................................... 189 20 Schwartz v. Hasty, 175 S.W.3d 621 (Ky. Ct. App. 2005)....................................................................... 202 21 Shulz v. Chadwell, 558 S.W.2d 183 (Ky. Ct. App. 1977)....................................................................... 202 22 Social Media Cases, 2025 WL 2807828 (Cal. Super. Sep. 22, 2025) .................................................... 189 23 Sonner v. Premier Nutrition Corp., 971 F.3d 834 (9th Cir. 2020) ......................................... 210, 211, 214 24 Spaw, LLC v. City of Annapolis, 156 A.3d 906 (Md. Ct. App. 2017) .................................................... 214 25 State ex rel. Dep't of Ecology v. Anderson, 620 P.2d 76 (Wash. 1980) ............................................ 213-14 26 State of Maryland v. Exxon Mobil Corp., 406 F. Supp. 3d 420 (D. Md. 2019)...................................... 6, 7 27 State of Wash. v. McKesson Corp. et al., No. 19-2-06975-9 (Sup. Ct. Wash. July 6, 2021).......... 210, 214 28 State v. Rouse, 647 N.W.2d 286 (Wis. Ct. App. 2002)........................................................................... 202 ix PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 11 of 235 1 Stollenwerk v. Tri-W. Health Care All., 254 F. App'x 664 (9th Cir. 2007) ............................................ 190 2 Stringer v. Nat'l Football League, 749 F. Supp. 2d 680 (S.D. Ohio 2009) ................................................ 9 3 Sullivan v. Daimler Trucks N. Am., LLC, 2019 WL 13563633 (D.S.C. Aug. 28, 2019) .......................... 11 4 Superior Farm Mgmt., L.L.C. v. Montgomery, 513 S.E.2d 215 (Ga. 1999) ........................................... 214 5 Teutscher v. Woodson, 835 F.3d 936 (9th Cir. 2016) ............................................................................. 211 6 Thacker v. Ethicon, Inc., 47 F.4th 451 (6th Cir. 2022) ........................................................................... 194 7 Torres v. Jai Dining Servs. (Phoenix) Inc., 497 P.3d 481 (Ariz. 2021).......................................... 187, 195 8 Under Armour, Inc. v. Ziger/Snead, LLP, 158 A.3d 1134 (Md. Ct. Spec. App. 2017) .......................... 203 9 Underwood v. Camden Cnty. Off. of Sheriff, 2024 WL 1366672 (D.N.J. Mar. 28, 2024) ..................... 189 10 United States v. Apex Oil Co., 579 F.3d 734 (7th Cir. 2009) ................................................................. 212 11 United States v. Philip Morris USA Inc., 449 F. Supp. 2d 1 (D.D.C. 2006) ....................................... 18-19 12 United States v. Price, 688 F.2d 204 (3d Cir. 1982) ............................................................................... 214 13 USA Truck, Inc. v. Sullair, LLC, 2020 WL 12656229 (M.D. Ga. Dec. 7, 2020)........................................ 9 14 Walton v. Premier Soccer Club, Inc., 334 A.3d 784 (Md. 2025) ........................................................... 194 15 Weinberger v. Romero-Barcelo, 456 U.S. 305 (1982) ........................................................................... 212 16 Wickersham v. Ford Motor Co., 853 S.E.2d 329 (S.C. 2020) ........................................................ 187, 196 17 Wilder v. Blue Ribbon Taxicab Corp., 719 S.E.2d 703 (S.C. Ct. App. 2011)………………………….205 18 Williams v. Schneider Elec. USA, Inc., 2023 WL 4374514 (Ky. Ct. App. July 7, 2023) ........................... 7 19 Wood v. Marathon Refining Logistics Serv. LLC, 2024 WL 2242688 (N.D. Cal. Mar. 21, 2024) ........ 211 20 Yonce v. SmithKline Beecham Clinical Lab'ys, Inc., 680 A.2d 569 (Md. Ct. Spec. App. 1996) ............ 190 21 Zakas v. Jackson, 835 S.E.2d 371 (Ga. Ct. App. 2019)………...………………………………………205 22 Zayo Grp., LLC v. Sols. Fiber Optic, Inc., 2025 WL 2157902 (E.D. Va. July 11, 2025) report and 23 recommendation adopted, 2025 WL 2155792 (E.D. Va. July 29, 2025) ........................................... 202 24 Statutes & Regulations 25 15 U.S.C. § 6501………………………………………………………………………………………..63 26 15 U.S.C. § 6502(b)(1)(A)(ii) ................................................................................................................. 171 27 15 U.S.C. § 6502(b)(1)(B) ...................................................................................................................... 171 28 x PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 12 of 235 1 18 U.S.C. § 2252…………………………………………………………………………………………69 2 16 C.F.R. § 312.4 .................................................................................................................................... 171 3 16 C.F.R. § 312.5 .............................................................................................................................. 63, 171 4 16 C.F.R. § 312.6 .............................................................................................................................. 63, 171 5 16 C.F.R. § 312.8 .................................................................................................................................... 171 6 Other Authorities 7 Brief for Common Sense Media and Frances Haugen as Amici Curiae Supporting Petitioners, 8 Gonzalez, et al., v. Google LLC, No. 21-1333 (Dec. 6, 2022)……………………………………….160 9 Az. Const. art. XI, § 1 et seq ................................................................................................................... 186 10 Ga. Const. art. VIII ................................................................................................................................. 186 11 Ky. Const. § 183 ..................................................................................................................................... 186 12 Md. Const. art. VIII, § 1 et seq. .............................................................................................................. 186 13 Fed. R. Civ. P. 56 ............................................................................................................................... passim 14 Fed. R. Evid. 702(a) .......................................................................................................................... 12, 205 15 Restatement (Second) of Torts § 433A ........................................................................................... 194, 200 16 Restatement (Second) of Torts § 821B ................................................................................................... 210 17 Restatement (Second) of Torts § 821C(2) .............................................................................................. 210 18 Restatement (Second) of Torts § 821F ................................................................................................... 187 19 Surgeon General's Warning on Social Media Platforms, N.Y. Times (June 17, 2024) …………………1 20 Murthy, Vivek H., Protecting Youth Mental Health: The U.S. Surgeon General's Advisory (2023) ...... 14 21 22 23 24 25 26 27 28 xi PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 13 of 235 1 I. INTRODUCTION AND "ROAD MAP" 2 American teens are facing an unprecedented mental health crisis, and Defendants' social media 3 platforms are a substantial contributing factor. As U.S. Surgeon General Vivek H. Murthy stated, "it is 4 time to require a surgeon general's warning label on social media platforms, stating that social media is 5 associated with significant mental health harms for adolescents." Vivek H. Murthy, It's Time to Require 6 a Surgeon General's Warning on Social Media Platforms, N.Y. Times (June 17, 2024). As the harmful 7 effects of Defendants' platforms have mounted, the Plaintiff school districts ("Districts") have been forced 8 to expand counseling, intervention, and mental health services, diverting limited resources from their core 9 mission of education. Defendants attempt to minimize this reality, placing the word "crisis" in quotation 10 marks, but the impact on the Districts is undeniable: they have been compelled to address student well- 11 being, at the expense of their core mission to educate. These costs are not static; recent survey data and 12 expert analyses demonstrate that the harms to student mental health—and the resulting burdens on school 13 systems—are ongoing and will continue into the future. 14 Each Defendant claims it owed no duty to the Districts, that the harms were unforeseeable, or that 15 federal immunity doctrines shield its inaction. The record proves otherwise. Defendants knew their 16 platforms' deliberate exploitation of children's attention for profit was fueling a youth mental health crisis 17 and destabilizing schools. Yet they concealed those facts and failed to warn either the Districts or students 18 and their parents. Defendants' internal research and independent studies leave no doubt: the harms to 19 youth mental health and school systems are measurable, persistent, and worsening—proof that these 20 outcomes were not only foreseeable but inevitable under Defendants' business model. 21 Defendants' conduct has substantially and foreseeably interfered with the right of children to 22 attend school in an environment conducive to learning and free from preventable psychological harm. The 23 resulting harms from this public nuisance—a widespread mental-health crisis, disrupted classrooms, and 24 the redirection of educational resources—are not isolated or accidental; they are the predictable outcome 25 of corporate choices to maximize youth engagement and advertising revenue at the expense of the Districts 26 and student well-being. 27 Defendants begin each of their motions by emphasizing the preexisting challenges school Districts 28 already face—poverty, racism, community violence, or natural disasters—as though those realities PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 14 of 235 1 absolve them of responsibility. But the law rejects that defense, recognizing that vulnerability does not 2 excuse misconduct; it magnifies its consequences. Every district endured COVID-19 and its disruptions, 3 but that shared challenge does not diminish Defendants' accountability for the distinct harms their 4 platforms imposed on already fragile educational systems. 5 This Omnibus Response provides the Court with a consolidated explanation of why Defendants' 6 overlapping motions fail, highlighting the common factual and legal disputes that must be resolved by a 7 finder of fact—disputes over duty, breach, causation, and damages that are the same for every District and 8 cannot be resolved on summary judgment. In addition to this filing, each of the six bellwether Districts 9 has submitted its own individual opposition applying their governing state law to their particular facts. 10 This filing is organized as follows: 11 Section II – Legal Standard: Establishes that under Rule 56, summary judgment is improper 12 where, as here, material facts are in dispute or credibility determinations are required. 13 Section III(A) – Liability: Makes clear that Defendants owed a duty of care to act reasonably to 14 prevent foreseeable harm to students and schools and breached that duty by 1) knowingly and substantially 15 causing and exacerbating the youth mental-health crisis, which the Districts are forced to address, and 2) 16 disrupting District operations and negatively impacting the school environment. The record also 17 establishes that Defendants' conduct created and maintained a public nuisance by unreasonably interfering 18 with the public rights to health, safety, and education. 19 Section III(B) – Causation: Explains that Defendants' "entanglement" arguments misconstrue 20 both law and fact. Plaintiffs are not required to isolate a single cause of harm; concurrent causes are for 21 the jury to resolve. And the record shows a direct nexus between Defendants' creation of a nuisance and 22 their negligent practices, omissions, and failures to warn, and the harms suffered by the Districts. 23 Section III(C) – Past Damages: Details the substantial evidence showing the Districts have 24 incurred compensable losses, including increased expenditures and diversion of school resources. Expert 25 analyses from Dr. Ward and Mr. Klein confirm that the value of teacher and staff time diverted to address 26 disruptions caused by Defendants' platforms represents a recognized and quantifiable injury under law. 27 Section III(D) – Future Damages: Establishes that the Districts' evidence of ongoing harm is 28 concrete and non-speculative. Future damages and abatement remedies are recognized and recoverable 2 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 15 of 235 1 under the laws of each jurisdiction. 2 Collectively, the six District filings and this Omnibus Response show that Defendants' platforms 3 have imposed real and ongoing harms on the Districts. The record—both shared and District-specific— 4 contains ample evidence of the creation of a public nuisance, and of duty, breach, causation, and damages. 5 Those factual disputes must be resolved by a jury, not on summary judgment. 6 II. LEGAL STANDARD 7 Summary judgment is appropriate only where there is no genuine dispute of material fact and the 8 movant is entitled to judgment as a matter of law. See Fed. R. Civ. P. 56(a); Celotex Corp. v. Catrett, 477 9 U.S. 317, 322 (1986). A dispute is "genuine" if a reasonable jury could return a verdict for the nonmoving 10 party, and "material" if it could affect the outcome of the case. Anderson v. Liberty Lobby, Inc., 477 U.S. 11 242, 248 (1986). When evaluating a motion for summary judgment, the Court must construe the evidence 12 and draw all reasonable inferences in favor of the non-movant. Matsushita Elec. Indus. Co. v. Zenith Radio 13 Corp., 475 U.S. 574 (1986); Brookside Assocs. v. Rifkin, 49 F.3d 490 (9th Cir. 1995); Ogden v. Bumble 14 Bee Foods, LLC, 2014 WL 27527, at 6 (N.D. Cal. Jan. 2, 2014). The Court may not weigh evidence or 15 make credibility determinations. Masson v. New Yorker Mag., Inc., 501 U.S. 496, 520 (1991) (citing 16 Anderson, 477 U.S. at 255); Atari Interactive, Inc. v. Redbubble, Inc., 515 F. Supp. 3d 1089, 1099 (N.D. 17 Cal. 2021). Because the Districts have presented ample evidence creating genuine disputes of material 18 fact, summary judgment must be denied. 19 III. ARGUMENT 20 A. Genuine material disputes of fact concerning Defendants' liability preclude summary judgment. 21 1. Defendants are liable to the Districts for failing to warn about the risks of 22 their platforms. 23 Defendants have fueled a youth mental health crisis by deliberately targeting school-aged children 24 with addictive social media platforms designed to maximize engagement and Defendants' profit. They 25 have also targeted schools, actively promoting student use of their platforms during the school day, while 26 failing to warn the Districts, their students, and parents about the serious mental health risks posed by their 27 platforms. The Districts have been left to bear the costs of the ensuing harms, including a compromised 28 educational environment and expenditure and diversion of resources to address student distraction and 3 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 16 of 235 1 mental health problems. 2 Defendants now seek to avoid responsibility for these harms, contending they owed no duty to the 3 Districts because the Districts were not direct users of their platforms. But the Court has already held that 4 Defendants do owe a duty to the Districts, and Defendants offer no basis to revisit that legal determination. 5 Defendants' attempt to relitigate this issue is as substantively baseless as it is procedurally improper. 6 Defendants owe a duty to warn the Districts of the harms caused by their deliberately addictive platforms, 7 as well as a duty to warn students and parents of the serious mental health risks their platforms pose. The 8 Districts have presented substantial evidence demonstrating they were harmed as a result of Defendants' 9 breach of those duties. Defendants cannot escape liability by disregarding the predictable impact of their 10 conduct on the very schools left to confront and manage the crisis they created. 11 a) The Court's prior rulings establish Defendants owed a duty to the Districts. 12 This Court held that Defendants owe the Districts a duty of care grounded in foreseeability, the 13 relationship between Defendants' conduct and the Districts' harms, and public policy. In re Soc. Media 14 Adolescent Addiction/Pers. Inj. Prods. Liab. Litig., 754 F. Supp. 3d 946, 972–78 (N.D. Cal. 2024) ("MTD 15 Order"). In denying Defendants' motion to dismiss the Districts' negligence claims, the Court concluded 16 that the Districts plausibly alleged a legal duty of care and expressly allowed their failure-to-warn theories 17 to proceed. Id. The Court explained that "three fundamental considerations" guide the duty analysis across 18 jurisdictions:1 "(1) the relationship between the parties, in particular, the relationship between the 19 defendant's conduct and the plaintiff's injury; (2) the foreseeability of the plaintiff's injury; and (3) public 20 policy concerns." Id. at 973. The Court found that all three factors supported a duty of care. Id. at 973–78. 21 Defendants now ignore these settled determinations. Instead, Defendants mischaracterize the 22 Court's ruling, incorrectly claiming it recognized merely a "general" duty, which excludes any duty to 23 warn of "financial harms" from student platform use. See, e.g., Defs.' Breathitt Mot. at 35. But the Court 24 expressly rejected Defendants' effort to dismiss failure-to-warn theories concerning "known risks of 25 26 27 1 Under Arizona law, foreseeability is not a factor in the duty analysis but the relationship between the 28 parties and public policy concerns are factors. Gipson v. Kasey, 150 P.3d 228, 231–34 (2007) 4 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 17 of 235 1 addiction attendant to any platform features or as to platform construction in general," and held that 2 Defendants owed the Districts a duty of care arising from their "intentional design choices to foster 3 compulsive use in their minor users and a failure to warn thereof." MTD Order, 754 F. Supp. 3d at 963– 4 64, 978. The Court further found that, "as [Defendants] became sophisticated in their targeting and capture 5 of minor users as to instill compulsive use of the platforms . . . it was objectively reasonable to expect the 6 alleged corollary injuries." Id. at 975. 7 This makes sense. The direct impact on the Districts includes financial harms, such as the 8 expenditure and diversion of resources to combat student use of Defendants' platforms during the school 9 day and to address the negative impacts of Defendants' platforms on students. These were not obscure, 10 unknowable consequences of Defendants' actions, but rather a natural result of Defendants' choice to 11 target schools and school-aged children and to design platforms promoting compulsive use, including 12 during school hours. See infra at § III.A.2. The Districts should have been warned about these harms. 13 Defendants' argument also ignores that if Defendants had warned the public of the potential negative 14 impacts on students—even without warning the Districts—the Districts could have taken steps to prevent 15 and mitigate the resulting financial harms they experienced. 16 Defendants cannot relitigate the Court's prior legal determinations at summary judgment. "[W]hen 17 a court decides upon a rule of law, that decision should continue to govern the same issues in subsequent 18 stages in the same case." Doe 1 v. Mayorkas, 530 F. Supp. 3d 893, 909 n.8 (N.D. Cal. 2021) (quoting 19 Musacchio v. United States, 577 U.S. 237, 244–45 (2016)). The Court's findings remain controlling—and 20 the record only strengthens them through evidence of Defendants' knowledge of youth engagement, 21 school-day disruption, and the foreseeable burdens their conduct imposed on schools. 22 b) Defendants owe a duty to warn the Districts directly. 23 Defendants owe a duty to warn the Districts as parties directly and foreseeably harmed by their 24 platforms' design and operation. Under well-established tort principles, a duty to warn arises when a 25 defendant's conduct creates a foreseeable risk of harm to others. MTD Order, 754 F. Supp. 3d at 973 (duty 26 arises where there is "sufficient juxtaposition of the parties in time and space to place the plaintiff in 27 danger from the defendant's acts") (quoting Quisenberry v. Huntington Ingalls Inc., 818 S.E.2d 805, 811 28 (Va. 2018)). This duty extends to all those the defendant should reasonably expect to be harmed by their 5 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 18 of 235 1 conduct. See State of Maryland v. Exxon Mobil Corp., 406 F. Supp. 3d 420, 463 (D. Md. 2019) ("duty to 2 warn extends . . . to third persons whom the supplier should expect to be endangered") (cleaned up). 3 Defendants' own internal research, marketing strategies, and school outreach confirm they knew 4 their platform designs were fostering compulsive use among students and producing widespread 5 consequences for schools. See infra §§ III.A.2.a.2-3, b.2-3, c.2-3, d.2-3 (discussing Defendants' targeting 6 of young people and schools); III.A.2.a.4, b.4, c.4, d.4 (discussing Defendants' knowledge of harms). 7 Internally, Meta researchers minced no words: "IG is a drug… We're basically pushers." Ex. 74 at 2152.2 8 "Teens are hooked despite how it makes them feel. Instagram is addictive, and time-spend on platform is 9 having a negative impact on mental health" Ex. 75 at 1428. Meanwhile, TikTok researched what "fuels 10 usage and adoption in High Schools." Ex. 423 at 7326. Snap acknowledged that "the ones that have the 11 Snapchat addiction have no room for anything else. Snap dominates their life." Ex. 922 at 4154. And 12 YouTube—despite knowing it "is the most widely used internet Platform by US teens"—recognized 13 "growing concerns that excessive screen time is stunting physical, emotional, and social development of 14 teens." Ex. 772 at 7914. 15 By deliberately targeting school-aged children and embedding features that promote compulsive 16 engagement, Defendants created and amplified foreseeable harms the Districts now must address every 17 day. Having created and profited from the very risks that injured the Districts, Defendants cannot now 18 escape liability by artificially narrowing the scope of their duty to warn by framing the Districts' claims 19 as a "novel" attempt to impose a duty on "indirectly injured" non-users of their platforms. See, e.g., Defs.' 20 Breathitt Mot. at 34-35. This Court has already rejected that idea. It is not novel to hold Defendants liable 21 for failing to warn those directly and foreseeably harmed by their conduct. See MTD Order, 754 F. Supp. 22 3d at 974 (holding "that 'the school Districts' injuries are sufficiently 'direct'" to impose a duty of care). 23 The Court's prior holding is squarely aligned with a long line of cases recognizing that those who create 24 foreseeable dangers must take reasonable steps to warn all who are likely to be affected—not merely end- 25 users. See, e.g., Dozier Crane & Mach., Inc. v. Gibson, 644 S.E.2d 333, 336 (Ga. App. 2007) ("The 26 common-law duty imposed upon suppliers of chattels includes the duty to warn of foreseeable dangers 27 2 28 All exhibits referenced in this brief are attached to the accompanying Declaration of Previn Warren. 6 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 19 of 235 1 arising from the reasonable use for which the product is intended and requires the exercise of reasonable 2 care to inform third persons of the dangerous condition or of the facts which make the product likely to 3 become dangerous."); State of Maryland v. Exxon Mobil Corp., 406 F. Supp. 3d 420, 499 (D. Md. 2019) 4 ("duty to warn extends not only to those for whose use the chattel is supplied but also to third persons 5 whom the supplier should expect to be endangered by its use") (cleaned up); New Jersey Dep't of Env't 6 Prot. v. E.I. du Pont de Nemours & Co., 2021 WL 6144081, at *7 (D.N.J. Dec. 30, 2021) ("New Jersey 7 courts have not limited the duty to warn to those with whom a defendant has a direct relationship"); 8 Williams v. Schneider Elec. USA, Inc., 2023 WL 4374514, at *4 (Ky. Ct. App. July 7, 2023) (unpublished) 9 ("Kentucky law is clear that duty is not predicated on classifications such as 'bystander or nonconsumer 10 or nonuser'" (quoting Jones v. Hutchinson Mfg., Inc., 502 S.W.2d 66, 69-70 (Ky. 1973)). 11 Unable to overcome this precedent, Defendants rely on a series of inapposite cases to 12 mischaracterize the Districts' claims as too attenuated to impose a duty. Defendants' reliance on Gourdine 13 v. Crews, 955 A.2d 769 (Md. 2008); CertainTeed Corp. v. Fletcher, 794 S.E.2d 641 (Ga. 2016); Quiroz 14 v. ALCOA Inc., 416 P.3d 824 (Ariz. 2018); and Kuciemba v. Victory Woodworks, 531 P. 3d 924 (Cal. 15 2023) is misplaced. Each involved remote, incidental injuries to individuals unconnected to the 16 defendants' actions. In Gourdine, the court declined to impose a duty on a drug manufacturer because the 17 injured motorist was a remote bystander, with no contact or connection to the manufacturer or product. 18 955 A.2d at 786 ("there was no contact between [the defendant] and [the decedent] whatsoever"). 19 Similarly, CertainTeed, Quiroz, and Kuciemba involved "take-home" exposure claims—arising from 20 asbestos and the coronavirus—where the plaintiffs were physically and relationally distant from 21 defendants' conduct. CertainTeed, 794 S.E.2d at 643; Quiroz, 416 P.3d at 827; Kuciemba, 531 P. 3d at 22 930-931. 23 Unlike the plaintiffs in those cases, the Districts are not remote from the conduct of Defendants. 24 Indeed, this Court has already found "sufficient juxtaposition of the parties in time and space to place the 25 [the Districts] in danger from the [Defendants'] acts." MTD Order, 754 F. Supp. 3d at 973–74 (citation 26 omitted). Defendants identify no facts that undermine this Court's prior holding. Their conduct was neither 27 passive nor attenuated: Defendants directly targeted schools and students, promoting youth engagement 28 during school hours while knowing the predictable consequences to the Districts. See infra §§ A.2.a.2-3, 7 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 20 of 235 1 b.2-3, c.2-3, d.2-3 (discussing each Defendant's targeting of school-aged children and schools); Ex. 313 2 at 5641 (Meta: "The Young Ones are the Best Ones"); Ex. 414 at 0624 (TikTok: high schooler[s] is a core 3 audience of our platform"); Ex. 824 (Snap "saw peaks of [Snapchat] activity during the school day"); Ex 4 733 at 8317 (YouTube: "Shorts is our big thing for teen appeal"). Their platforms deliberately captured 5 minors' attention throughout the school day, making schools a direct and foreseeable locus of harm—not 6 incidental bystanders. See id.; Kennedy Krieger Inst. v. Partlow, 191 A.3d 425, 457 (Md. 7 2018)(distinguishing Gourdine and recognizing that a duty may arise where there is a "relational link and 8 direct contact" between the defendant and the injured party). By deliberately fostering compulsive use 9 among students, and failing to warn the Districts, students, and parents of the risks of their platforms, 10 Defendants directly and foreseeably caused the Districts' injuries. MTD Order, 754 F. Supp. 3d at 968, 11 974–76; see also In re JUUL Labs, Inc., Mktg., Sales Practices & Prods. Liab. Litig., 497 F. Supp. 3d 12 552, 645-651 (N.D. Cal. 2020) (finding that "youth-oriented advertising" and "targeting of youth" created 13 a foreseeable risk of harm to school Districts forced to respond to vaping on campus). 14 Further, this Court has already rejected Defendants' floodgates argument, finding "[t]he doomsday 15 view is a dramatic overstatement." MTD Order, 754 F. Supp. 3d at 976. The Court held that the recognized 16 duty is "no more expansive than defendants' own intentional and targeted actions to specific schools." Id. 17 at 976–77. Thus, imposing duty here does not expand tort law. It enforces accountability for a defined 18 category of misconduct: targeting minors through harmful platform design and failing to warn of the 19 resulting risks. See In re JUUL Labs, 497 F. Supp. 3d at 658 (finding that defendants violated Arizona 20 law and public policy "by marketing an addictive product to school-age children, leading to an epidemic 21 of youth vaping and corresponding harms to schools"). As this Court previously recognized, "defendants' 22 concrete and particularized awareness of potential and actual harms to [the Districts] appropriately tailors 23 the scope of duty." MTD Order, 754 F. Supp. 3d at 976-977. 24 c) Defendants owe a duty to warn students and parents about their 25 platforms' dangers. 26 As part of their duty to the Districts, Defendants were obligated to warn the Districts' students and 27 their guardians of the risks posed by their platforms. Instead, Defendants deliberately fostered compulsive 28 use of their platforms, "which foreseeably caused the plaintiff school Districts to respond by expending 8 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 21 of 235 1 resources to mitigate the impact of such use in their schools." MTD Order, 754 F. Supp. 3d at 968, 974– 2 76; see Ex. 38 at 8062 (discussing Meta's goal to "saturate school networks"); Ex. 537 at 3040–41 3 ("[TikTok] is addictive," and "compulsive usage correlates with a slew of negative mental effects"); id. 4 at 3041 ("compulsive usage on TikTok is 'rampant'"); Ex. 930 at 5370 (Snapchat's "addictiveness" affects 5 school performance); Ex. 701 at 1255-56 (discussing YouTube's goal to "[i]ncrease habitual users" and 6 "focus on making YouTube a daily habit"). 7 Courts have long recognized that a defendant may be held liable for failing to warn a non-plaintiff 8 when that failure leads to a plaintiff's injury. See Parris v. 3M Co., 595 F. Supp. 3d 1288, 1337–38 (N.D. 9 Ga. 2022) (holding that PFAS manufacturers' duty to a proposed class of landowners included a duty to 10 warn a downstream mill of the dangers posed by PFAS); USA Truck, Inc. v. Sullair, LLC, 2020 WL 11 12656229, at *3 (M.D. Ga. Dec. 7, 2020) ("the violation of a duty to exercise reasonable care toward one 12 person to whom that duty is clearly owed can proximately cause an injury to someone other than the 13 person to whom the duty is directly owed"); see also Stringer v. Nat'l Football League, 749 F. Supp. 2d 14 680, 703 (S.D. Ohio 2009) ("If the manufacturer's failure to warn influenced the conduct of a third party, 15 and that third party's acts or omissions were the proximate cause of the plaintiff's injury, then the 16 manufacturer may be held liable."). 17 Even though Defendants are the movants and bear the burden on summary judgment, they present 18 no evidence that they warned the Districts' students and their parents about the risks associated with 19 Defendants' platforms. Nor do they contest that the Districts' injuries were a foreseeable result of this 20 failure to warn. That alone demonstrates a genuine issue of fact for a jury. Instead, Defendants invoke 21 three inapposite cases—CertainTeed, Gourdine, and Quiroz—to argue for a rule this Court has already 22 rejected. That reliance is misplaced and contrary to established precedent as well as the law of the case. 23 Multiple courts—including Defendants' authority CertainTeed in its concurrence—have refused 24 to adopt Defendants' broad reading. CertainTeed turned not on any categorical limitation on the duty to 25 warn, but on the absence of foreseeability due to the parties' highly attenuated relationship. USA Truck, 26 2020 WL 12656229, at *3. As the court in USA Truck explained, "Certainteed did not hold that a 27 manufacturer could not be held liable when a manufacturer breaches its duty to warn a user and that breach 28 proximately harms a third party." 2020 WL 12656229, at *3; see also Parris, 595 F. Supp. 3d at 1337 9 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 22 of 235 1 (holding the reasoning in CertainTeed "does not foreclose a duty to warn" non-plaintiffs); CertainTeed, 2 794 S.E.2d at 646 (noting that the majority opinion "does not address whether [a plaintiff] might have a 3 claim as an alleged injured party resulting from the failure to warn [an end user.]") (Melton, J. concurring). 4 No such attenuation exists here. 5 The same is true of Gourdine and Quiroz, both of which turned on the absence of any meaningful 6 relationship between the defendant's conduct and the plaintiff's harm. In Gourdine, the court found no 7 duty where "there was no contact between [the manufacturer] and [the decedent] whatsoever." 955 A.2d 8 at 786. Likewise, Quiroz emphasized that "duties are relational" and held that a company's connection to 9 an employee's family member was too attenuated to support a duty. 416 P.3d at 839. In each case, the 10 courts declined to impose a duty because the plaintiffs were remote from the defendants' conduct. 11 The factors that defeated liability in CertainTeed, Gourdine, and Quiroz are not present. As this 12 Court has held, the relationship between Defendants and the Districts is sufficiently direct, given that 13 Defendants' platforms were directed at students and schools; the Districts' injuries were foreseeable; and 14 imposing a duty to warn accords with established public policy. MTD Order, 754 F. Supp. 3d at 972–78. 15 d) Defendants' failure to warn caused the Districts' injuries. 16 Contrary to Defendants' assertions, genuine issues of material fact exist as to whether Defendants' 17 failure to warn caused the Districts' injuries. The District-specific oppositions set forth in detail the 18 substantial record evidence demonstrating Defendants failed to warn the Districts, their students, or 19 guardians, and the direct injuries the Districts suffered as a result. See Districts' MSJ Opps.; Ex. 1085; Ex. 20 1086; Ex. 1087; Ex. 1088; Ex. 1089.; Ex. 1090; Ex. 1083; Ex. 1106; Ex. 1107; Ex 1108; Ex. 1109; Ex. 21 1110; supra § III.A.4. As Defendants' own authority demonstrates, this is sufficient evidence of causation 22 to proceed to trial. See Jones by & through Jones v. IC Bus, LLC, 626 S.W.3d 661, 682 (Ky. Ct. App. 23 2020) (finding directed verdict improper where there was evidence demonstrating defendant had 24 knowledge of the dangerous condition and failed to provide any warning); see also R & R Insulation Servs. 25 v. Royal Indem. Co., 705 S.E. 2d 223, 235 (Ga. Ct. App. 2010) (finding summary judgment should be 26 denied where there was evidence that defendants caused injury, because "[w]hile the jury may ultimately 27 determine that any additional warning would not have prevented the harm here, that question is not for 28 this Court to determine at this stage of the proceedings."). Unable to contend with this record, Defendants 10 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 23 of 235 1 suggest the Districts lack evidence that is not required under the law. 2 Defendants rely on a single inapposite case to argue expert testimony is required on the "adequacy 3 and efficacy" of warnings here. Defs.' Breathitt Mot. at 36; Defs.' Charleston Mot. at 41; Defs.' Harford 4 Mot. at 43-44; Defs.' Irvington Mot. at 34; Defs.' Tucson Mot. at 47 (citing Nelson v. Am. Honda Motor 5 Co., 2021 WL 2877919 (W.D. Pa. May 17, 2021)). They are wrong. In Nelson, the defendant had issued 6 warnings, and the court required expert testimony to evaluate "the contents of the existing warnings." Id. 7 at *7 (emphasis added). Other cases cited by Defendants likewise confirm that consideration about the 8 adequacy of warnings is reserved for cases where there is an existing warning. Allen v. Long Mfg. NC, 9 505 S.E.2d 354, 360 (S.C. Ct. App. 1998) (evidence that existing warning was inadequate created issue 10 of fact as to whether an adequate warning could have changed plaintiff's conduct); Hickerson v. Yamaha 11 Motor Corp., U.S.A., 2016 WL 4367141, at *5 (D.S.C. Aug. 16, 2016) (plaintiff failed to show existing 12 warning inadequate where plaintiff testified she did not read it); Georgia Cas. & Sur. Co. v. Salter's Indus. 13 Serv., 734 S.E.2d 415, 421 (Ga. Ct. App. 2012) (no evidence existing warning inadequate). 14 Here, Defendants do not dispute that they provided no warnings about the risks their platforms 15 pose. Because there are no warnings to evaluate, expert testimony on their "adequacy" or "efficacy" is 16 unnecessary and would be illogical. Courts repeatedly recognize that juries are capable of understanding 17 a failure-to-warn claim where no warning exists. See Bullock v. Volkswagen Grp. of Am., Inc., 107 F. 18 Supp. 3d 1305, 1316 (M.D. Ga. 2015) ("Plaintiffs contend that Defendants knew of risks with the design 19 but provided no warning. A jury can understand this issue without a warnings expert."); Sullivan v. 20 Daimler Trucks N. Am., LLC, 2019 WL 13563633, at *7 (D.S.C. Aug. 28, 2019) ("Plaintiff is not required 21 to produce a human factors expert to testify about the need for a warning about a dangerous condition 22 when no warning exists at all"); Ruggiero v. Yamaha Motor Corp., U.S.A., 2017 WL 1197755, at *10 23 (D.N.J. Mar. 31, 2017), aff'd, 778 F. App'x 88 (3d Cir. 2019) (expert testimony on warnings not necessary 24 where "contents of [existing] warnings are not at issue"); Hanna v. Ward Manuf., 2016 WL 3196467, at 25 *2 (M.D. Fla. June 9, 2016) ("[T]he Court agrees with Plaintiffs' argument that, [where no warnings were 26 provided], expert testimony was not required to prove that Defendant's warnings were inadequate."); see 27 also Fed. R. Evid. 702(a) (expert testimony must assist the jury in understanding evidence or determining 28 a fact at issue). 11 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 24 of 235 1 Even if expert testimony were required, the record is replete with such testimony supporting the 2 Districts' claims. Indeed, Districts' experts explain that Defendants' platforms pose unreasonable risks to 3 children, that Defendants failed to warn anyone of those risks, and that this failure amplified the platforms' 4 harmful effects, which the Districts are now forced to confront. For example, Dr. Seth Noar's expert report 5 details the specific, foreseeable hazards Defendants failed to disclose, and explains how their omissions 6 fell below well-established principles and standards for warnings. Ex. 997 ¶¶ 214–48. As Dr. Noar further 7 opines, Defendants' failure to adopt and enforce a proper, standards-based warning was a driving force 8 behind the harms children suffer from Defendants' platforms. Id. ¶¶ 214–64. Likewise, Tim Estes explains 9 that Defendants could have provided clear and effective warnings about the risks of their platforms; 10 without them schools and educators cannot effectively "deal with the fall-out from the negative mental 11 health effects of the platforms." Ex. 989 ¶ 316. Defendants also could have, but chose not to, provide 12 warnings in Congressional testimony and/or in materials they distributed to schools, like their guides for 13 parents. See infra § III.A.2.a.6, b.6, c.6-7, d.6-7; see, e.g., at 358:14-386:6 ( 14 testifying that CEO's congressional testimony failed to include warnings to 15 students, parents and schools regarding the negative consequences of Defendants' platforms); Ex. 16 at 281:3 (same TikTok employee testifying that the TikTok Guide for Parents, which TikTok distributed 17 to school PTAs and directly to schools, failed to warn of harms caused by Defendants' platforms). 18 2. Defendants breached their duty of care. 19 a) Meta's Conduct 20 (1) Meta's business model 21 Meta is an extraordinarily large and well-capitalized company.3 Ex. 1 at 15-17 (over $164 billion 22 3 For purposes of this and the Districts' individual opposition briefs, "Meta" refers to Meta Platforms, Inc. 23 ("MPI") and Instagram, LLC. Meta argues that only its parent company MPI "is responsible for the 24 development and operation" of Instagram. See Defs.' Breathitt Mot. at 39. But there is a genuine dispute of fact about this, precluding summary judgment. Instagram's "Terms of Use" state that Instagram is 25 "owned or controlled by Instagram, LLC," Ex. 558 at 8784; requires consumer disputes to be directed to "Instagram, LLC," id. at 8788-89; and does not so much as mention MPI. Further, evidence reveals that 26 Instagram, LLC knew about harms on the platform: Instagram, LLC contracted with a third-party research group for a quantitative study, which found that 52% of teens had negative experiences on Instagram. Exs. 27 1113-1115. As for the remaining named Meta subsidiaries and affiliates, the Districts and Meta are 28 negotiating a stipulation pursuant to which they will be dismissed from the Districts' actions. 12 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 25 of 235 1 in revenue and over $5 billion in dividends and dividend equivalents in 2024). It has earned substantially 2 all those riches from selling advertising on its platforms. Ex. 6 at 31:10-13, Ex. 7 at 338:25-339:12, Ex. 8 3 at 24:3-7, 248:13-249:6. However, Meta is fundamentally different than traditional forms of ad-supported 4 media like cable TV and broadcast radio. An ad on TV is played to everyone in the same local market. 5 Not so for Meta's platforms. Because Meta ingests an enormous amount of data about its users, it can sell 6 advertisers the ability to "target" specific ads to specific users. Ex. 12 at 31:1-32:32; see Ex. 2 at 77:24- 7 79:21 (describing types of data collected about users). 8 Meta has a dark history of exploiting user information to target ads. See Ex. 8 at 65:14-20 (former 9 Meta advertising executive: "I helped to build an advertising system that provably, provably, could change 10 the way that you feel about a brand of toothpaste, or any other product."). At his deposition, former 11 Facebook research scientist (and current member of the U.K. Parliament) Dr. Joshua Simons recalled a 12 conversation about one disturbing feature of the ad delivery system: "[T]he system knows … you're a 13 new mum because of all the data that it has about you, and it was showing … women who just had children 14 ads for things that would sort of make them feel bad about their body in some way, you know, like body 15 cream or body adjustment surgery or whatever." Ex. 13 at 63:23–65:23. He explained how a colleague 16 had worked on ways to change this but was "unable to get those things rolled out, because… they would 17 result in fewer ads being clicked on and therefore less revenue generated by the ad delivery system." Id. 18 This abuse of user data for ad targeting is consistent with the Congressional testimony of former 19 employee Sarah Wynn-Williams, who stated that Meta packages "sensitive personal data together and 20 sells access to it so that advertisers can … tailor their messages to resonate with people experiencing 21 emotional vulnerabilities or desires that an advertiser finds useful for pitching products." Ex. 202 at 3; see 22 Ex. 221 at 9638-39 (Meta "marketing science" presentation examining "moments when young people 23 need a confidence boost"). Discovery in this litigation has revealed that the source code of Meta's ranking 24 algorithm includes the variables "Had eating disorder" and "Had weight change." Ex. 14 at 461:5-465:12. 25 Meta's proposed algorithm expert could not rule out the possibility that Meta infers these attributes of its 26 users and uses this information to target ads. Id. 27 Fundamentally, Meta's business model means that it makes more money the more time users spend 28 on its apps. Ex. 8 at 93:14-94:4 ("More usage correlates with more money."); Ex. 9A at 94:5-16; Ex. 6 at 13 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 26 of 235 1 368:21-369:13. Meta is therefore incentivized to "maximiz[e] the amount of time and attention that people 2 kind of dedicate to the platform, thus allowing for … more space and time to present the advertising." Ex. 3 12 at 31:8-32:10. 4 Testimony from multiple deponents confirms that Meta can and does attempt to increase the 5 amount of time users spend on Instagram and Facebook to make more money. Former senior UX 6 researcher explained: "It seemed like when I was there, that Meta was trying to increase 7 engagement, trying to increase how much time people spent using their apps in an effort to have more 8 eyeballs on the apps for a longer period of time, which can get them more advertising revenue." Ex. 10 at 9 31:5-32:14. Dr. Simons observed that bonuses "were awarded to those teams who significantly boosted 10 engagement," including "the amount that a user interacts on the platform, the time they spend on it, [and] 11 the frequency … with which they visit the platform," all of which "produces more ad revenue." Ex. 13 at 12 134:24-13. Even Meta's since-withdrawn algorithm expert, Dr. Larry Birnbaum, conceded that Meta sets 13 as a "goal metric" how much money it can earn from every minute someone spends on Facebook, a 14 measure it internally calls "monetizable user value." Ex. 14 at 307:16-309:8; Ex. 15 at slide 22. 15 By targeting users' fundamental need for social connection—a survival instinct—and weaponizing 16 the mathematical structure of social networks, Meta makes it nearly impossible for users to stop using 17 their platform on their quest to monopolize their time and generate ad revenue. 18 (2) Meta's targeting of school-aged children 19 Unfortunately for a generation of young people, Meta—following the same playbook once used 20 by Big Tobacco—determined years ago that the most valuable users, meaning those who generate the 21 most advertising revenue, are the youngest. Ex. 313 at 5641 ("The Young Ones are the Best Ones"); see 22 also Vivek H. Murthy, Protecting Youth Mental Health: The U.S. Surgeon General's Advisory (2023) 23 (social media companies have "adopted tactics reminiscent of the tobacco industry"). Meta "hypothesized, 24 that social media could be somewhat similar to, say, music where you, you know, find your tastes early 25 on and stick with them over time." Ex. 7B at 570:9-571:16. The data bore this out. Users who join 26 Facebook as 13-year-olds "have an LTV [lifetime value] of approximately $350, higher than all other teen 27 ages (13-18), due to their very high long-term retention." Ex. 16 at 9025; Ex. 21 at 118:23-119:13, 176:13- 28 178:12; Ex. 24 at 1 ("The difference in long term retention can be dramatic, with the people who signed 14 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 27 of 235 1 up for Facebook when they were young teens being 100% higher"). Teenage users presented another 2 advantage: They could serve as gateways into the "household ecosystem," including pre-teen siblings. Ex. 3 18 at 1070 ("Teens strongly influenced preteens' understanding of what and how frequently to share on 4 Instagram"); Ex. 22 at 252:18-253:9 (Instagram researcher: "teens are often the ones that other members 5 of the household learn about not only Instagram, but social media in general through."). 6 Armed with these insights, Meta concluded it should "acquire teens as young as possible." Ex. 16 7 at 9025; see also Ex. 28 at 6914 (2017: "Teens: Let's always build and prioritize with them in mind."); 8 Ex. 6 at 114:16-17, 114:22-115:3 (Instagram CEO: "Mark [Zuckerberg] is suggesting that teen time spent 9 be our top goal in 2017."); Ex. 26 at 5014 (2018: "Winning Teens = Winning Generations."); Ex. 17 at 10 0071 (2021: "Youth and Teens are critically important to Instagram."); Ex. 18 at 1071 ("capturing the teen 11 user cohort on IG is critical."); Ex. 33 at 11 (2024: "Acquiring new teen users is mission critical to the 12 success of Instagram."). 13 Despite its relentless "focus[] on the teen retention and growth," Ex. 20B at 642:19-643:6, Meta's 14 leadership knew by 2015 that the company was losing its youngest users—and treated that decline as an 15 existential threat. Internal reports warned of a "significant DAP and engagement problem with teens," 16 calling it "worse than we previously thought and getting worse over time." Ex. 25 at 7; see also Ex. 9 at 17 202:18-205:7; Ex. 26 at 4998 (2018: "Teen Opportunity Cost… total present value loss ~$12 billion"); 18 Ex. 27 at 8621 ("our brand isn't faring well with teens"); Ex. 31 at 40 ("Instagram has an existential crisis 19 with [t]eens"); Ex. 50 at 5556 ("We have a teens problem"); Ex. 20A at 241:16-242:14 ("Alarm bells 20 started going off as youth ceased using the platform or engaging the platform as much"); Ex. 9 at 198:12- 21 20 (addressing teen market "definitely one of Mark [Zuckerberg]'s priorities and concerns."); Ex. 6 at 22 126:23-127:10 (Instagram CEO Adam Mosseri: "if US teens were using Instagram less, that would be a 23 big concern… It would be a big risk"). 24 Rather than address the safety or well-being of these users, Meta chose to double down— 25 knowingly sacrificing safety to recapture its most profitable audience. The company rushed to deploy new 26 features designed to hook teens, in line with its internal mantra to "Move Fast and Break Things." Ex. 8 27 at 282:12–283:3 (this phrase "literally was a poster on the walls."). At Zuckerberg's direction, employees 28 undertook a "lockdown sprint" to launch Facebook Live in early 2016 as "the beachhead we need to 15 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 28 of 235 1 expand into other use cases in video and teens," the "two spaces we care about deeply but are behind in." 2 Ex. 32 at 9332–33; Ex. 11A at 337:8–21. Meta intentionally omitted parental and teacher safeguards, with 3 Zuckerberg warning that notifying adults "will probably ruin the product from the start" and instructing 4 that the company "be very good about not notifying parents/teachers." Ex. 205 at 6655. Worse still, Meta 5 learned that Live was being used to broadcast teen suicides and attempts, an issue it did nothing to prevent 6 or warn about before the feature's launch. Ex. 8 at 285:10-287:13. As one former Meta VP conceded, "it 7 was a growth imperative to make Live Video a product and release it, as opposed to slowing down and 8 really trying to think through all the negative ways it could be used." Ex. 8 at 287:15-25. Only in April 9 2025—nearly a decade later—did Meta add "new built-in restrictions…so that teens under 16 can't access 10 Live without a parent's permission." Ex. 271 at 1. 11 Similarly, when the U.S. launch of TikTok in 2018 caused a "decline in US Teen growth"—which 12 Meta perceived as "an existential threat for not just IG but the company overall"—it responded by rushing 13 to release Reels, a feature that knocked off TikTok's endless feed of vertical, short-form videos. Ex. 30 at 14 0933; Ex. 3 at 31:7-10, 31:23-32:4 ("There were clear market competitive motivations to build that 15 product fast"); 215:21-216:9 (describing "increased emphasis on pushing Reels" and "build[ing] this 16 product as engagingly as possible"). The launch was a success, in the sense that millions of young users 17 began using Reels in a matter of months. Ex. 3 at 64:11-15; see also Ex. 29 at 0126 ("Reels has grown 18 and TikTok decelerated in 2022, and so we have closed a significant amount of the gap, but we are still 19 well behind."). However, Reels was also rushed out without appropriate safety protections for young 20 users. Ex. 2A at 38:20–24 (Q: "In your experience, was the launch of Reels on Instagram handled in a 21 way that best protected the safety of Instagram's youngest users?" A: "No, I don't think so."); Ex. 3 at 22 53:11-21 ("the integrity guardrails that they put in place were not sufficient"). And Meta continued to 23 launch changes to Reels they knew made it less safe. Ex. 206 at slide 54 ("We released a high volume 24 (100+ in H1 2021) of product launches, 17 leading to integrity regressions."). Meta did not pause to 25 evaluate whether those features could operate together to present even greater safety risks. Ex. 3 at 50:9- 26 12, 52:3-53:9 (Meta "did not test for whether there were interaction effects" but "did see…later on, the 27 cumulation of all of these launches, did lead to integrity regressions"). Meta failed to warn the Districts 28 or the public about any of these dangers. Ex. 2A at 56:12-57:17; Ex. 3 at 86:11-19; Ex. 997 ¶¶ 142-148. 16 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 29 of 235 1 (3) Meta's targeting of schools 2 Meta's obsession with dominating the time and attention of youth led it to embark on a concerted, 3 years-long campaign to embed Instagram and Facebook directly into school communities. See Ex. 373 at 4 0761 ("one of the things we need to optimize for is sneaking a look at your phone under your desk in the 5 middle of Chemistry :)"); Ex. 37 at 4706 ("For U.S. Teens, we should focus Facebook on high school 6 communal activities…tipping schools may be high impact"), 4714 (2016: "per-high school adoption is a 7 crucial driver of teen Facebook engagement"); Ex. 44 at 0316 ("there's an opportunity to do classical FB 8 growth on a high-school-by-high-school basis"); Ex. 36 at 3595 (2018: "Winning schools is the way to 9 win with teens because an individual teen's engagement is highly correlated with school MAP [monthly 10 active persons] penetration…Winning Teens = Winning High Schools."); Ex. 38 at 8062 (goal: "saturate 11 school networks"); Ex. 45 at 3732 ("Over an 18 month period we expect to drive an estimated 5% teen 12 DAP (+2.5% teen MAP) from our high school work."); Ex. 46 ("[e]ngaging the vast majority of teens in 13 an area / school with our products is crucial to driving overall time spent in the same area"); Ex. 53 at 14 7145 (2023: "Can we take this a step further and use school networks as a lever for acquisition?"), 7146 15 (2023: "How can we position Instagram as integral to navigating school relationships, especially during 16 transition periods?"); Ex. 55 at 5749 (2024: "we will need to invest in virality within school communities 17 to get teens and their friends on IG."). The clear goal of Meta's campaign was to legitimize its platforms 18 in the eyes of teachers, administrators, and parents, so that their use among students would become 19 normalized. 20 Meta's school initiative took a number of different forms. First, Meta developed the technical 21 capability to determine when teen users were at school. Ex. 14 at 468:14-469:5 (Meta's expert 22 acknowledging that this capability is embedded in the source code of Meta's ranking algorithm); Ex. 39 23 at 145:3-146:16 (Meta's corporate representative acknowledging that Meta is able "to look at different 24 signals to understand when teens are online and jump to whether that is during the school day or not, or if 25 they are physically at school"). Meta could have, but did not, use this information to reduce teens' 26 engagement with their platforms during school hours. Ex. 14 at 470:5-471:17. 27 Second, Meta embarked on efforts to identify students' specific school affiliations—even though 28 half of teens decided against sharing that information with the company. Ex. 39 at 94:10-21 ("a lot of 17 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 30 of 235 1 teens would put Hogwarts"), 97:19-4 ("only 9.3 million out of the 18.7 million active …included a 2 school"). This has included efforts to infer students' school affiliations using complex data mapping 3 techniques. Ex. 1179 at 7260 ("it is essential that we correctly identify the school a given teen is 4 attending"), 7265 ("we plan to embed each teen as a data point and each school as a multidimensional 5 gaussian" which "will allow us to directly derive the probability of attending any given school"); Ex. 43 6 at 2355 (describing how a teen's location can be mapped to a school, if analyzed "during school hours"); 7 Ex. 49 at 1566 (acknowledging "FB ML [machine learning] school inference model"). It also includes a 8 recent effort (dubbed "Project Gryffindor") through which Meta encourages teens to display their school 9 affiliations on Instagram. Ex. 42; see Ex. 40 at slides 8-9 (discussing need to acquire "databases of US 10 high schools" for this project). Meta has taken these measures because "increasing high school profile 11 completion drives engagement and DAP [Daily Active Persons]." Ex. 41 at 4558. Indeed, Meta has 12 estimated that "[u]nlocking new networks through school communities" has the "potential to drive ~150K 13 DM [Direct Message] Teen DAU [Daily Active Users]." Ex. 55 at 5739-5740; see Ex. 36; Ex. 45 at 3731; 14 Ex. 46. 15 Third, using location and school-specific information collected through data harvesting, Meta 16 targeted push notifications to students within specific schools—which it called "school blasts." Ex. 54 at 17 9227. The express aim of these "blasts" was to "tip schools from inactive to active via network effects and 18 incrementally increase teen MAP [monthly active persons] in those schools." Ex. 54 at 9227. 19 Fourth, Meta paid the National PTA and Scholastic to conduct an extensive outreach campaign 20 targeting schools and families. Meta chose these partners precisely because they were "trusted, respected 21 organizations" with "significant credibility" that could "be a public validator" for the company and "get 22 our materials into the hands of parents, grandparents, and educators at scale." Ex. 48 at 4519. Meta's 23 agreements with these organizations "include significant website placement, social media promotion, 24 targeted physical distribution, community activations, and post-campaign surveys." Id.; Ex. 47 at 4527. 25 Meta's conduct mirrors the tactics of Big Tobacco, which used "youth smoking prevention" and school- 26 based programs to infiltrate classrooms and legitimize its products under the guise of education—a 27 strategy condemned in United States v. Philip Morris USA Inc., 449 F. Supp. 2d 1, 853–71 (D.D.C. 2006) 28 18 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 31 of 235 1 (finding tobacco companies intentionally marketed to youth through schools while falsely claiming to 2 promote public health). 3 Meta "began a relationship with the National PTA as early as 2014, possibly earlier." Ex. 39 at 4 25:6-26:3. In 2018 alone, Meta paid the PTA $110,000, with part of that budget "coming from marketing." 5 Ex. 48 at 4519. In exchange, Instagram was made an "Official partner for PTA's first-ever Back to School 6 Week campaign," given a "Dedicated Back to School landing page housing Instagram-supplied 7 educational content," and benefitted from an "Email blast[s]" to "~16K PTA presidents at middle and high 8 schools nationwide" and "~150k PTA members." Ex. 48 at 4520. Internally, Meta praised the partnership 9 for granting access to the PTA's "16.2 million families across the U.S.," noting that "If you want to 10 connect with parents at scale, you work with the PTA." Id. Through its PTA partnership, Instagram 11 distributed parents' guides, hosted focus groups with educators and administrators, and held events at 12 "nearly 200 schools across the country, in every single state." Ex. 39 at 190:11-192:20; see also Ex. 51 at 13 7467 (planning conversation with PTA on "specific sponsorship amount" and "parent roundtables"), Ex. 14 52 at 6709 (discussing "Facebook PTA Digital Toolkit Campaign"). Meta even worked with the PTA to 15 "[c]apture footage of PTA families trying out" new Facebook products "and offering testimonials"— to 16 "build trust with parents and with press." Ex. 277 at 5130. 17 Meta's paid relationship with Scholastic mirrored its partnership with the National PTA in both 18 scale and purpose. Ex. 56 at 6341 ("we worked with Scholastic in a big way for our last IG parents' push"). 19 In 2018, Instagram's marketing team paid Scholastic $135,000 and, in exchange, received "Physical 20 distribution" of its materials "to 20,000 classrooms across the U.S. (Reach: 20,000 teachers / 600,000 21 students/ families assuming 30 students per class)." Ex. 48 at 4519. Meta recognized that Scholastic's 22 reputation would amplify its message, noting that "98% of teachers feel confident about sharing 23 information that they receive from Scholastic with parents of their students." Ex. 48 at 4520. As Meta 24 explained internally, "Scholastic gives us the ability to win educators and the school community, creating 25 a ripple effect to allow us to win parents and families. Educators and schools don't feel like they have 26 enough information to help teens in their schools; a Scholastic partnership helps fix that." Id.; see also Ex. 27 50 at 5583 ("Leverage established relationship with Scholastic to facilitate activations"). 28 19 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 32 of 235 1 Fifth, Meta engaged in a deliberate campaign of direct outreach to schools and school districts to 2 normalize its products under the guise of education and safety—again mirroring the tactics once used by 3 Big Tobacco to gain access to youth audiences. The company presented "safety roadshows" at high 4 schools across the country, including Evergreen High School and Miami-Dade County Public Schools. 5 Ex. 57 at 3853; Ex. 58 at 3188. It hosted "Teacher Institutes" for 400 educators at its offices (Ex. 64 at 6 8612); addressed "THE conference for charter schools around the US." in partnership with the National 7 Alliance of Charter Schools (Ex. 65 at 0703; Ex. 59 at 4307); and created a "digital citizenship" program 8 "[d]esigned with teachers in mind," piloted in multiple middle schools (Ex. 60 at slides 3, 7; Ex. 61). Meta 9 launched a Facebook-powered "personalized learning platform" in select charter schools, through Summit 10 Public Schools. Ex. 62 at 4392. It piloted an experimental teen-only app in "the 10 largest schools in NYC 11 and Miami." Ex. 63 at 5481. Even after this lawsuit was initiated, its efforts have continued; in 2024, 12 Instagram launched what it calls a "School Partnership Program." Ex. 39 at 39:3-10, 40:12-14, 18-21; see 13 also Ex. 73 at 1695 (Instagram head of policy noting "a lot of conversations lately with Principals at 14 schools and school districts"). 15 Sixth, Meta undertook marketing efforts specifically designed to reach teenagers in their role as 16 students. This included a "Back-to-School with IG" brand campaign that "showcase[d] teens navigating 17 the transition between summer vacation and back-to-school through product-centric storytelling." Ex. 66 18 at 2285; see also Ex. 67 at 6488 ("launching a new Back-to-School hub"). Meta's outreach went further: 19 the company recruited and paid 13–17-year-old "teen tastemakers to act as our plug at local high schools 20 within 5 key markets." Ex. 68 at 6448; see Ex. 71 at 0328 ("US Teen Lifetime Value" presentation asking, 21 "What's a reasonable level of funding for a teen ambassador program?"). These "ambassadors" received 22 compensation—including Amazon gift cards, branded "swag" like hats and sticker sheets—and were 23 required to sign nondisclosure agreements. Ex. 68 at 6453-54, 6449. Meta used these teen ambassadors to 24 "build excitement at local high schools" for products such as its "high school directory" pilot (Ex. 69 at 25 0219, 0221, 0241) and Reels (Ex. 70 at 9347). 26 As the foregoing should make clear, Meta had every conceivable opportunity to put school 27 administrators, teachers, parents, and students on notice if it knew about safety and health risks associated 28 with students' use of Instagram and Facebook. It had every conceivable opportunity to warn schools if it 20 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 33 of 235 1 knew those risks could materialize in the classroom. Meta did not do so. Instead, it used paid partnerships 2 with validators and generic educational toolkits to infiltrate school communities, legitimize Facebook and 3 Instagram, and promote platform adoption (including by schools themselves)—with the predicted and 4 desired result of growing teen engagement to drive long-term revenue. 5 (4) Meta's knowledge of harms 6 What Meta knew—and what it failed to reveal—should shock the conscience. Internal studies 7 conducted by Meta, both qualitative and quantitative, repeatedly confirmed that Instagram and Facebook 8 use could lead to a panoply of negative outcomes for teens, including addiction, sleep disruption, anxiety, 9 depression, negative appearance comparison, and body image problems. 10 We begin with addiction. And we start with what is perhaps the most succinct acknowledgment of 11 the problem, from a chat between two UX Researchers: "oh my gosh yall IG is a drug…We're basically 12 pushers… We are causing Reward Deficit Disorder bc people are binging on IG so much they can't feel 13 reward anymore…like their reward tolerance is so high…I know Adam [Mosseri] doesn't want to hear it 14 – he freaked out when I talked about dopamine in my teen fundamentals leads review but its undeniable! 15 Its biological and psychological….the top down directives drive it all towards making sure people keep 16 coming back for more. That would be fine if its productive but most of the time it isn't…the majority is 17 just mindless scrolling and ads." Ex. 74 at 2152-53. 18 Far from being one-off musings, these insights are echoed in a chorus of evidence from other Meta 19 researchers. See Ex. 75 at 1420 ("Teens talk of Instagram in terms of an 'addicts narrative' spending too 20 much time indulging in a compulsive behaviour that they know is negative but feel powerless to resist."), 21 1428 ("Teens are hooked despite how it makes them feel. Instagram is addictive, and time-spend on 22 platform is having a negative impact on mental health."); Ex. 76 at 3172 ("56% of IG teens surveyed say 23 it's difficult to manage how much time they spend on social media and 14% say IG makes it worse"); Ex. 24 85 at 5016 ("Teens may sometimes be using social media more than they intended to without paying 25 attention to it, such as engaging in habitual use....Teens may also believe that it's challenging to limit their 26 use or feel like it's hard to exert self-control"); Ex. 19 at 77:24-78:6, 80:15-23, 128:19-129:3 ("It wasn't 27 that teens didn't like it. It's that they didn't think that it was good for them. The teens that we talked to … 28 thought that they should be controlling it. They didn't think that they could."); Ex. 78 at 7036 ("(1) teens 21 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 34 of 235 1 feel addicted to IG and feel a pressure to be present, (2) like addicts, they feel that they are unable to stop 2 themselves from being on IG, and (3) the tools we currently have aren't effective at limiting their time on 3 the ap[p]"); Ex. 13 at 67:2-10 ("addiction was one of the types of harm…that I talked about with senior 4 executives"); Ex. 77 at 3673 ("sophisticated reinforcement learning might increase the problem of tech 5 addiction in Stories, News Feed, or Instagram"); Ex. 79 at 4205-4206 ("Are there teens that use too much 6 Instagram? It sure looks like it….It seems relatively clear that younger people are less equipped to handle 7 social media addictions….In the US alone, there are 550K+ teens that spend 28 hours a week on Instagram 8 (4 hours a day, 7 days)."); Ex. 110 at 3026 ("it is clear that clinicians are seeing people reporting problems 9 with Facebook use."); Ex. 132 at slide 32 ("social media addiction: heavy usage of Facebook correlates 10 with lower happiness"). 11 Meta's understanding of the addiction issue dates at least back to the summer of 2017, when a data 12 science intern internally published a research note titled, "Have we made people addicted to Facebook?" 13 Ex. 104 at 0086; see Ex. 140. As part of this research, the intern and his supervisors spoke to the author 14 of a book called Hooked to "discuss[] his extensive research into creating habit-forming products[.]" Id. 15 The team then fielded a survey, received responses from 1,300 Facebook users, built a model, and 16 determined that "we could call addicted…nearly 200k people." Ex. 104 at 0089. The intern emailed Adam 17 Mosseri—then the Head of News Feed at Facebook, now the CEO of Instagram—asking to discuss his 18 findings. Mr. Mosseri blew him off. Ex. 6 at 391:15-23 ("can't meet with every intern that reaches out."). 19 Just weeks later, two of Meta's former senior executives offered scathing critiques of Facebook's 20 addictiveness. On November 9, 2017, Sean Parker, "the second president" of Facebook, Ex. 11A at 48:3- 21 14, made public statements that "[t]he thought process that went into" building Facebook "was all about 22 how we consume as much of your time and conscious attention as possible. And that means that we need 23 to sort of give you a little dopamine hit every once in a while because someone liked or commented on a 24 photo or post or whatever. And that's going to get you to contribute more content…it's a social validation 25 feedback loop…it's me, it's Mark…[we] understood this consciously, and we did it anyway." Ex. 134. 26 "God only knows what it's doing to our children's brains." Ex. 135. At his deposition, Mark Zuckerberg 27 credited Mr. Parker as "a smart person" who "helped out a lot with the business" and who influenced his 28 thinking around "business strategy." Ex. 11A at 48:15-49:18, 52:3-8; id. at 51:10-18 & Ex. 136 22 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 35 of 235 1 ("massively influential on my thinking"). Nonetheless—even though Mr. Zuckerberg "saw the headlines" 2 around Mr. Parker's comments—he did not reach out to him to discuss or "look into anything based on 3 this." Id. at 55:2-19, 55:25-56:6 ("I don't need to investigate that."). 4 On November 13, 2017, another Facebook executive went public with his concerns—Chamath 5 Palihapitiya, a vice president at Facebook from 2007 to 2011 with whom Mr. Zuckerberg was in regular 6 communication during his tenure. Ex. 11A at 93:24-94:15. Mr. Palihapitiya stated that "The short-term, 7 dopamine-driven feedback loops that we have created are destroying how society works" and "eroding 8 the core foundations of how people behave." Ex. 137. "We curate our lives around this perceived sense of 9 perfection because we get rewarded in these short-term signals, hearts, likes, thumbs up, and we conflate 10 that with value….And, instead, what it really is, is fake, brittle popularity that's short-term and that leaves 11 you even more…vacant and empty before you did it. Because then it forces you into this vicious cycle 12 where you're, like, what's the next thing I need to do now, because I need it back." Ex. 137. After the 13 press reported on his comments, Mr. Palihapitiya wrote to Mr. Zuckerberg apologizing for "all the pickup 14 today" (but conspicuously not disavowing the substance of his remarks). Ex. 138 at 6409-10. Mr. 15 Zuckerberg did not respond to the email and has never spoken to Mr. Palihapitiya since. Ex. 11A at 16 103:23-104:7, 106:12-17 & Ex. 1181. 17 While Meta's leadership ignored growing evidence of a problem, researchers at the company 18 began to undertake a series of qualitative, quantitative, longitudinal, and even experimental studies 19 confirming the widespread phenomenon of what it euphemistically called "problematic use." See Ex. 78 20 at 7036-37 ("Do we want to call it addiction?" "we would never want to say that!"); but see Ex. 107 at 21 0181 (acknowledging that problematic use "is sometimes referred to as 'social media addiction' 22 externally."). Meta internally defined problematic use as "experiencing both of the following issues either 23 'very often' or 'all the time': • Lack of control or feelings of guilt over Facebook use. • Negative impact 24 in at least one of the following areas: sleep, parenting, social relationships, or productivity." Ex. 80 at 25 5381. "Behaviors associated with PU [problematic use]" include "Frequent checking," "Overall amount 26 of time spent," and "Passive late night use." Ex. 76 at 3173; see also Ex. 109 at slide 16-17 (several 27 behaviors (e.g. total time spend, late night use, video chaining) and surfaces (e.g. Explore, Reels, 28 notifications) "associated with low perceived control on IG."). 23 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 36 of 235 1 In 2018, Meta's researchers "paired a survey of 20,000 U.S. Facebook users measuring perceptions 2 of problematic use with demographic and behavioral data for the prior month." Ex. 83 at 0059. That 3 study—whose results were shared directly with Mark Zuckerberg and Meta COO Sheryl Sandberg— 4 provided a "deep understanding" that the prevalence of problematic use among U.S. Facebook users is 5 "55% mild, 3.1% severe." Ex. 81 at 0761 (email to Mark Zuckerberg), Ex. 82 (attachment); see Ex. 11A 6 at 274:24-275:6 (Zuckerberg: "3 percent of billions of people is a lot of people…it's millions of people."). 7 Meta never publicly disclosed the true results of this study. It did, however, publish research falsely 8 stating, "we estimate (as an upper bound) that 3.1% of Facebook users in the US experience problematic 9 use." Ex. 97 at 2 (emphasis added); see Ex. 11B at 680:9-683:8. 10 In 2019, Meta's researchers "extended our measurement framework" and "revised our survey 11 questions" to "follow the same format … as is standard practice in mental health assessment scales." Ex. 12 99 at 0168. Using this framework, assessed against a "large, multi-national dataset" of 42,000 participants, 13 Meta determined that "problematic use appears to affect 12.5% of people" and "[a]nother 11% of people 14 report negative life impact 'very often' but [report] only 'sometimes' lacking control." Ex. 99 at 0169-70; 15 see Ex. 80 at 5381 ("12.5% of people worldwide experience problematic use"). Meta has never publicly 16 disclosed the results from this study. 17 At the end of 2019, a lead UX researcher at Facebook completed a 24-person qualitative study, 18 consisting of "1:1 interviews with 'problematic' Facebook users." Ex. 107 at 0178. Participants described 19 their addiction in unsparing terms: "The time I spend is not healthy, it's like an addiction. Opening the 20 app every half hour, it's not healthy." Ex. 107 at 0196. "When you have a beer on the weekend, you drink 21 it. You have to. It's the same thing with Facebook." Id. "[Y]ou can get into this rabbit hole and might not 22 get out." Id. "I spend a lot of time on the videos because they start automatically." Ex. 107 at 0201. "People 23 liking things can be addictive. I feel compelled to see who liked it. I think it's a bad habit because [I'm] 24 always checking." Id. "I'm on Facebook everyday, every moment. Literally, every moment; just not when 25 I'm in the shower. It's a habit…I lose the notion of time." Ex. 107 at 0202. While some participants 26 attempted "fruitless" self-remediation tactics to alleviate these problems, researchers reported that "none 27 had found a full solution" and most resorted to "deleting the [Facebook] app or deactivating at least once 28 to curb use." Ex. 107 at 0204. Meta did not disclose these findings to the public. 24 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 37 of 235 1 Around the same time, a senior UX research leader at Instagram (and a PhD in social psychology, 2 Ex. 11A at 83:6-19) prepared a work plan that included a recommendation that Meta warn the public about 3 its research findings: "Because our product exploits weaknesses in the human psychology to promote 4 product engagement and time spent [and that because of our reliance on AI, we cannot currently control 5 for when content is being strategically leveraged to manipulate the opinions and moods of individual 6 people] if we are committed to the wellbeing of individuals, we need to a) alert people to the effect that 7 the product has on their brain (for example: your view of yourself is affected by other people's highlight 8 reel, etc) and b) provide them with tools to have more control over their own experience." Ex. 105 at 4015 9 (brackets in original). Meta did not follow this recommendation and did not disclose the known effects of 10 its products on the brain. Ex. 11A at 87:2-93:1. 11 In addition to studying problematic use, Meta conducted parallel analyses on user well-being. In 12 2019, the company surveyed 30,000 Instagram users—once in May and again in October—to measure the 13 reach and intensity of perceived negative experiences on the platform. Ex. 100 (May 2019 results); Ex. 14 101 (Oct. 2019 results); Ex. 102 at 36:25-37:10 ("We were aiming to measure subjectively bad 15 experiences."). The findings were alarming: "the perceived reach of negative experiences on Instagram is 16 quite high," Ex. 102 at 44:23-45:14; Ex. 101 at slide 6, and "teens have significantly more negative 17 experiences than adults … for 11 out of the 18 problem areas that we asked about," Ex. 101 at 7. 18 Meta also conducted an analysis of 13- to 17-year-olds that "was meant to be a cross section of 19 Instagram users" and "[a] representative sample to the best of our ability." Ex. 19 at 103:10-13, 104:4-15; 20 Ex. 98 at 9175 (methodology included survey of 2,503 users). The results were stark: "One in five teens 21 say that Instagram makes them feel worse about themselves." Ex. 98 at 9193. Of those surveyed, "39% of 22 teens who had felt that they needed to create a perfect image said that that feeling started on Instagram," 23 Ex. 19 at 106:20-24; "41% of teens who felt not attractive said that that feeling started on Instagram," id. 24 at 106:25-107:3; "24% of teens who felt that they were not good enough said that that feeling started on 25 Instagram," id. at 107:13-16; "21% of teens who said they felt alone or lonely said that that started on 26 Instagram," id. at 107:25-108:3; "10% of teens who said they were down, sad or depressed said that that 27 feeling start[ed] on Instagram," id. at 108:4-7; "6% of teens who said that they felt that they wanted to kill 28 25 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 38 of 235 1 themselves said that that feeling started on Instagram," id. at 108:11-14; and "9% of teens who said that 2 they felt that they wanted to hurt themselves said that it started on Instagram," id. at 108:15-18. 3 Other internal studies conducted during the same period confirmed these patterns. Roughly one- 4 third of teens who reported problematic social media use said that Instagram made that experience worse. 5 Ex. 102 at 151:14–152:6; Ex. 123 at slide 19. In another survey of 1,000 teens, 23% said they "waste too 6 much time" on Instagram, and 19% said they "miss out on life" often or very often because of it. Ex. 111 7 at 32; Ex. 19 at 58:6–25. Taken together, these studies left Meta with unmistakable evidence that its 8 platforms were worsening teen well-being. 9 Meta did not voluntarily disclose the results of these 2019 qualitative surveys. The findings became 10 public only after a whistleblower leaked them to Congress and the press in 2021, prompting widespread 11 outrage. Ex. 23 at 125:11–126:6; Ex. 209 at 9310. In response, Meta released "annotated" versions of the 12 presentations designed to downplay and obscure their conclusions. Privately, the authors of this research 13 (Meta's own employees) were unsparing in their assessment of this move. See Ex. 102 at 209:23-300:4 14 (author of May and October 2019 presentations: "they're absolutely throwing us under the bus and making 15 us sound incompetent"), 301:5-8 ("they're basically trying to discredit the work so much that it's like it 16 didn't get leaked at all."); Ex. 1182 at 1964 ("I understand why they have to defend against the causal 17 claim externally, but internally why is it so hard to take responsibility for shitty experiences that users 18 have on the platform that we absolutely could do more to limit? …I think their perception is that the worst 19 thing here is the leak itself, followed by the language/mischaracterization of the article, followed by the 20 actual problem facing humans and the fact that we haven't devoted enough resources to it."); Ex. 19 at 21 307:6-21 (author of teen survey work: "it makes it sound like I was just guessing"), 308:12-15 ("that 22 phrasing of maybe researcher speculation. Like, that is the part that sat very unwell with me then and kind 23 of still does now."); Ex. 103 at 5013 (co-author of research: "like PLEASE just focus on what they are 24 REALLY criticizing you about and acknowledge it and stop throwing your researchers under the bus"). 25 Meta's research into user harm extended beyond qualitative studies. In late 2019, it "designed an 26 experimental deactivation study, in which we will randomly ask some people to stop using Facebook and 27 Instagram for a month (compared to a group who will continue to use as normal), helping us explore the 28 impact that our apps have on polarization, news consumption, well-being, and daily social interactions. 26 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 39 of 235 1 We'll partner with Nielsen and use a combination of (a) surveys, (b) log data from Facebook and 2 Instagram, and (c) usage data from smartphones." Ex. 87 at 5647; Ex. 88 at slides 2-5 (study design); Ex. 3 89 (Nielsen contract); Ex. 92 at 2432 (codename "Project Mercury"). Meta undertook this in-house 4 experimental study to fix what it perceived as shortcomings in the external literature. Ex. 96 at 0466 ("our 5 design is of much higher quality"); Ex. 87 at 5647 ("one of our first causal approaches to understand the 6 impact that Facebook has on people's lives….Everyone involved in the project has a PhD"). 7 To Meta's disappointment, pilot tests of the deactivation study confirmed that "[p]eople who 8 stopped using Facebook for a week reported lower feelings of depression, anxiety, loneliness, and social 9 comparison." Ex. 90 at slides 2, 6-7. Rather than investigate further, or sound the alarm, Meta halted the 10 project—claiming that participants' feedback was biased by "the result of the existing media narrative 11 around the company." Ex. 93 at 0429; Ex. 94 at 3905. But see Ex. 93 at 0429 (Nick Clegg: "is it really 12 possible to claim that the effect is all due to media conditioning/perception?"); Ex. 95 at 1927 ("the nielsen 13 study does show causal impact on social comparison "). 14 One Meta employee warned, "if the results are bad and we don't publish and they leak, is it going 15 to look like tobacco companies doing research and knowing cigs were bad and then keeping that info to 16 themselves? I went…oh." Ex. 91 at 1943. That is exactly what Meta did. The company never publicly 17 disclosed the results of its deactivation study. Instead, Meta lied to Congress about what it knew. Compare 18 Ex. 105 at 4007 (2017: listing "Increase in anxiety" among the "Known Negative Effects of FB and/or 19 Social Media in General on Teens"); Ex. 98 at 9196 (2019: "young people openly attribute their increased 20 level of anxiety and depression to Instagram"); Ex. 90 at slide 2 (Sept. 2020: cessation of Facebook use 21 leads to lower depression and anxiety) with Ex. 106 at 124-125 (Dec. 2020 response to questions from 22 U.S. Senate: Q: "Is Facebook able to determine whether increased use of its platform among teenage girls 23 has any correlation with increased signs of depression within this demographic…?" A: "No." Q: "Is 24 Facebook able to determine whether increased use of its platform among teenage girls has any correlation 25 with increased signs of anxiety within this demographic…?" A: "No."); Ex. 11 at 238:20-242:19. 26 Meta's internal research, and its external deception, continued. In December 2020, a member of 27 Meta's core data science team (who holds a PhD in economics, Ex. 141 at 2) authored an internal study 28 acknowledging "[t]here are reasons to worry about self-control and use of our products" and presenting a 27 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 40 of 235 1 "quick rundown of evidence"—including "[a]n experiment [which] found that a 1-month break from 2 Facebook improved self-reported wellbeing." Ex. 108 at 5595 & Ex. 139 at 0119. In response, another 3 senior data scientist at Meta (who also holds a PhD in neuroscience, Ex. 142 at 2, and taught a university 4 course on addiction, Ex. 143 at 1) warned: "It seems clear from what's presented here that some of our 5 users are addicted to our products. And I worry that driving sessions incentivizes us to make our product 6 more addictive, without providing much more value. How to keep someone returning over and over to the 7 same behavior each day? Intermittent rewards are most effective (think slot machines) reinforcing 8 behaviors that become especially hard to extinguish—even when they provide little reward, or cease 9 providing reward at all." Ex. 139 at 0127; see also Ex. 13 at 67:12-17. 10 Meta never publicly acknowledged these findings from its own in-house addiction specialists. 11 Instead—just weeks later—Mark Zuckerberg testified before Congress and denied that Meta profits from 12 creating addictive products. Ex. 144; Ex. 11A at 154:16–155:11. This deliberate contradiction between 13 Meta's internal science and its public testimony exemplifies a broader pattern: the company knew its 14 platforms fostered addiction and psychological harm but chose to deny, obscure, and mislead. 15 In 2022, Meta conducted a "mixed method study," surveying and interviewing over 1,000 "mental 16 health clinicians (including psychiatrists, psychologists, therapists, social workers)." Ex. 128 at slides 5- 17 6. Eligible clinicians had "2+ years experience post-licensure" and "Provided care for at least 30 patients 18 in the past 3 months." Id. This study revealed that "the majority of clinicians believe that social media can 19 be addictive," with fully 85% of U.S. clinicians endorsing this proposition. Id. at slide 9. 20 In 2023—"following resolution of a senior leadership-level legal escalation"—Meta's researchers 21 launched a longitudinal study to follow 4,000 U.S. teenagers over the academic year. Ex. 84 at 7080 22 ("Meta & Youth Social Emotional Trends" study); see also Ex. 86 ("I really prefer not to have a cute 23 acronym (MYST). That makes things sound more interesting when/if it leaks."). The goal of this survey 24 was to measure teens' attentiveness ("Possible unintended, unaware, and/or habitual use") and capability 25 ("Perceived self-efficacy to exert control, if desired"). Ex. 85 at 5018-19. Early results were troubling: 26 more than "one-third (37%)" of teens reported low attentiveness on Instagram, while slightly less than "1 27 in 3 teens (28%) reported experiencing challenges with capability." Ex. 85 at 5015, 5019. The research 28 also confirmed that parents and family circumstances are not to blame for this problem. Ex. 85 at 5021 28 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 41 of 235 1 ("Parental and household factors have little association with teens' reported levels of attentiveness to their 2 social media use….family income, parental marital status, and other household characteristics…were not 3 associated with the teens' feeling of attentiveness."), 5026 ( "There is no association between either 4 parental reports or teen reports of parental digital caregiving/supervision and teens' survey measures of 5 attentiveness or capability."). Meta has never disclosed the results from this study. 6 That same year, other researchers conducted a 12-country survey of 30,000 adults and 30,000 teens 7 (13-17) on Facebook and Instagram, to identify "on-platform behaviors associated with perceived low 8 control" over platform usage. Ex. 109 at slide 2, 10. Consistent with the MYST study, nearly 40% of youth 9 reported "low attentiveness" and 20% of youth reported "low capability" on Instagram. Ex. 109 at slide 10 13; Ex. 11A at 227:11-21; see also Ex. 982 ¶ 372 ("Three to six hours of social media at ages 13-16 was 11 associated with a 60% increased relative risk of internalizing problems at ages 14-17 and greater than 6 12 hours per day was associated with a 78% increased risk."). This research went on to assess how such 13 compulsive use interfered with daily life—specifically, whether it prevented individuals "from fulfilling 14 [their] responsibilities in other areas of [their] life (for example, work, school, or relationships)." Ex. 109 15 at 9. The results were stark: users with "low attentiveness" reported 20–25% higher life interference, and 16 those with "low capability" reported 17–23% higher interference. Ex. 109 at slide 13. Meta has never 17 disclosed these results. 18 As the foregoing study demonstrates, Meta understands not only that its platforms are addictive to 19 teens, but also that this addiction produces a cascade of related mental health harms. In 2020, an Instagram 20 researcher (who holds a doctorate in public health, Ex. 113A at 328:10-329:6) developed a "conceptual 21 model of adolescent social media behaviors and mental health," in collaboration with his colleagues. Ex. 22 112 at 2-3. The model shows that problematic social media use is linked to poor sleep, low self-esteem, 23 negative body image, and mental health challenges—each of which fuels further problematic use, creating 24 a self-reinforcing cycle. Ex. 112 at 3; Ex. 113A at 341:20–343:21 ("there's a feedback loop with mental 25 health challenges"). It also identifies adolescents with "existing vulnerabilities" as particularly at risk, 26 whether those vulnerabilities operate independently or in combination. Ex. 113A at 337:1-15. 27 28 29 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 42 of 235 1 2 3 4 5 6 7 8 9 10 11 12 Ex. 112 at 3. 13 To be clear, the connection between problematic use and mental health issues for teenage users 14 was not a one-off observation by a single researcher. Far from it. For example, Meta knew that addictive 15 use of Instagram and Facebook could cause disrupted sleep and poorer mental health for young users. Ex. 16 114 at 9142 ("Nighttime social media use is associated with poorer mental health in teens due to displacing 17 sleep; this is the most direct relationship between social media use and teen well-being. The most 18 vulnerable teens offline are also the most vulnerable teens online"); Ex. 115 at 2192 ("For late night use, 19 the negative impacts, such as on work/school performance and mood, or little or poor sleep (especially on 20 young people) are well-documented."); Ex. 116 at 10 (43.3% of teens have at least one session on 21 Instagram between 12am-4am each week and 4.1% of teens "spend 18+ min on IG" during these hours), 22 15 (two thirds of teens who exhibit problematic late-night use "start a session from a notification"); Ex. 23 117 at 8 (2018 feedback from "US teen": "When I'm sleeping, I get a lot of Instagram notifications"); Ex. 24 118 at 9909 ("As session count increases, more and more people report feeling like they spend a lot of 25 time on Facebook or report sleep problems."). Despite this extensive body of evidence, Meta remained 26 silent, failing to warn the Districts and the public. Ex. 997 ¶¶ 142–148. 27 28 30 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 43 of 235 1 As discussed above, Meta also knew that addictive use of its platforms contributed to depression 2 and anxiety. Ex. 90 at slide 2; Ex. 105 at 4007; Ex. 98 at 9196. Internal research captured this vividly: 3 teens described "The Me on Instagram" as "one-dimensional, boxed in (small box), not in control, dark, 4 alone, pulled down, low esteem, rabbit hole, anxious, overthinking." Ex. 119 at 1117; see also Ex. 19 at 5 85:7-12 ("some teens have told us that they feel anxiety around their Instagram usage"). Meta's 2022 6 survey of 1,000 mental health clinicians confirmed what its own researchers already knew—81% of 7 clinicians said social media exacerbated patients' anxiety disorders, and 78% said it worsened depressive 8 disorders. Id. at slide 24. Despite this extensive evidence, Meta concealed the truth from the public and 9 lied to Congress about what it knew. Ex. 106 at 124-125. 10 Over years of internal research, Meta repeatedly found that Instagram use drove negative social 11 comparison and eroded teen self-esteem. Ex. 81 at 0761 (email to Mark Zuckerberg: "deep understanding" 12 of "negative drivers that occur frequently on FB" including "Social comparison (prevalence: 40% mild, 13 5% severe)"); Ex. 127 at 8 (internal survey of 5,793 randomly selected participants in 2018 revealed that 14 "33% of people have been feeling worse about themselves on IG for 'several months to a year'), 24 15 ("regression…shows that age and gender are the strongest predictors of negative social comparison"); Ex. 16 111 at 32 (32% of surveyed teens felt "like I have to look perfect" on Instagram); Ex. 19 at 58:6-25; Ex. 17 119 at 1098, 1101 ("Social comparison is worse on Instagram"; "66% of teen girls on IG experience 18 negative social comparison"; "32% of teen girls said that when they felt bad about their bodies, Instagram 19 made them feel worse"); Ex. 120 at 8700 (internal survey of over 50,000 users finding that "One-quarter 20 of people on IG (23%) feel a lot or extreme pressure to look perfect on Instagram, and one-third of teen 21 girls (34%) do," and that "Nearly half of teen girls (48%) often or always compare their appearances on 22 IG"), 8704 ("Appearance-based comparison on Instagram is common and can have significant 23 implications for individuals"); Ex. 102 at 153:16-24 ("somewhere between, you know, 15 and 20%" of 24 respondents said that Instagram made social comparison worse); Ex. 128 at slide 42 (mixed-methods study 25 of 1,000 clinicians: "Encouraging negative comparison was most cited as negative impact on mental health 26 patients"); Ex. 129 at 1890, 1895 ("Results from a 100k-person survey + behavioral data" finding that 27 "About 1 out of 10 people experience negative social comparison on Instagram often or always") & Ex. 28 130 at 217:22-218:2 (this is millions of people). Despite this extensive body of evidence, Meta never 31 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 44 of 235 1 warned the public. Ex. 997 ¶¶ 142–148. Instead, it continued to promote Instagram as a positive and 2 empowering space for teens—knowing full well that its own research showed the opposite. 3 Nor did Meta disclose its knowledge of a related harm: that addictive use of Instagram fuels body 4 image issues for teens. Ex. 121 at 5133 ("research suggests a causal link between Instagram use and 5 appearance comparison and/or body image issues."); Ex. 131 at 6265 ("external research consistently 6 points to appearance comparison as a key step towards internalization of the 'thin ideal' and other beauty 7 ideals which often lead to body image issues and for some eating disorders/ suicide/ self injury."); Ex. 123 8 at 5021 (survey of 22,410 users revealing that "we make body image issues worse for 1 in 3 teen girls" 9 and that "[m]ore teen girls thought that IG made body image issues worse") & Ex. 122A at 204:5-18, 10 214:21-215:2; Ex. 124 at 1221 ("the team's vision should be to move IG out of the position of being 11 uniquely worst for teens' social comparison and body image") & Ex. 20B at 502:5-503:21; Ex. 125 at 12 0058 ("it seems patently obvious that phone usage, selfie-culture, and photo-editing are contributing to 13 anxiety, depression, and body dysmorphia….we have to assume we are contributors / amplifiers."); Ex. 14 126 at 7066 ("there is substantial evidence to suggest that Instagram and Facebook use can increase body 15 dissatisfaction."). Meta did not warn the public, including the Districts, about these problems. 16 Finally, Meta knew that addictive use of its platforms could lead to self-injury, including eating 17 disorders and suicidality. Ex. 173 at slide 7 (discussing how "addiction" on its platforms "can lead to more 18 downstream effects like self-injury or suicide" and "eating disorders"); Ex. 132 at slide 30 ("model for 19 how social media usage may be linked to suicide/self-harm"); Ex. 128 at slide 24 (showing that over 30% 20 of clinicians believed that social media "had a negative role" in suicidal behavior disorder and non-suicidal 21 self-injury disorder). This is particularly concerning since Meta also knew that teens represented the "vast 22 majority of users who admit and promote self-harm and suicide are teens (45%-60%)" and "reporters of 23 self-harm and suicide are also disproportionately represented by teens (52%-56%)" despite "only 24 mak[ing] up 18% of DAP." Ex. 172 at 0010. Meta did not warn anyone about these problems. 25 In short, Meta understood that use of Instagram and Facebook could lead to serious, negative 26 mental health harms—particularly for teenage users. Ex. 114 at 9141 ("best research suggests that 27 Facebook is net bad for well-being"); Ex. 108 at 5595 & Ex. 133 at 71:2-72:15 (internal study stating, 28 "Higher Facebook use is correlated with worse psychological states (wellbeing, loneliness, etc.)"); Ex. 32 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 45 of 235 1 119 at 1124 ("Users experience of downward spiral is exacerbated by our platform"), 1125 ("Aspects of 2 Instagram exacerbate each other to create a perfect storm"), 1126 ("Mental health outcomes related to this 3 can be severe" and include "depression," "body dysmorphia," and "eating disorders"); Ex. 19 at 165:10- 4 166:9 ("teen girls that we talked to very clearly associated this spiral that they were experiencing on 5 Instagram with some, you know, pretty real – pretty bad outcomes."), 167:19-25 ("the teen girls and young 6 adult women that we talked to felt as if Instagram was exacerbating these issues for them"). 7 Meta also knew or should have known that the harms caused by its platforms would inevitably 8 spill into schools, where those teens spend much of their lives and receive education, mental health, and 9 other services. And, as discussed above, Meta had countless opportunities to warn the Districts through 10 its partnerships with the National PTA, Scholastic, and its extensive outreach to schools. See supra 11 § III.A.2.a. 4; see also, e.g., Ex. 169 at 4962 ("A Parent's Guide to Instagram"); Ex. 170 at 2240 ("National 12 PTA and Facebook Launch Digital Families Community Events"); Ex. 148 at 0829 ("Communications 13 Plan: Back to School 2020…showcase how Facebook uses technology to … connect parents, students and 14 teachers"). It said nothing. As one former Meta executive put it, "[t]his move was part of a broader pattern: 15 when transparency threatens profits, Meta chooses secrecy." Ex. 8 at 122:22–123:12; id. at 105:12–107:19 16 ("The company doesn't provide any ability for us as a society to know"). By ignoring these warnings, 17 Meta allowed its products to infiltrate classrooms, disrupt learning environments, and contribute directly 18 to the youth mental health crisis now overwhelming schools nationwide. 19 (5) Meta's failure to exercise reasonable care 20 Meta argues that some of its users are, at most, addicted to and hurt by the content on Instagram 21 and Facebook, not by the platforms themselves. That is a convenient defense, invoking Section 230. But 22 it is not true. And there is substantial evidence, coming again from Meta's internal documents and 23 research, proving that point. See e.g., Ex. 466 at 3 ("our results cannot be attributed solely to the content 24 of the post"). In this section, we turn to the dangers and harms inherent in Meta's platforms, Meta's failure 25 to warn Districts about those risks, and its consistent prioritization of its profits over the remediation of 26 known dangers to kids, including the Districts' students. 27 28 33 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 46 of 235 1 Before addressing those issues as to specific "platform features," we first take up Meta's failure to 2 warn as to the risks "attendant to … platform construction in general." Dkt. 1267 at 14. Again, internal 3 Meta documents demonstrate that its platforms as a whole are designed in unsafe ways that create harms 4 including, but not limited to, addiction. Ex. 105 at 4015 ("Product features that are designed to exploit 5 insecurity, or provide a dopamine rush (likes, notification, the pull-down-to-see, the infinite scroll, etc), 6 to increase time spent, are inherently at odds with well-being and take away from people's ability to 7 consciously focus[] on activities that add value to their lives."); Ex. 80 at 5384 ("Triggers" for "loss of 8 control over time spent on Facebook" include "Notifications," "auto-play," "Fear of missing out 9 (FOMO)," and the "Mystery of the algorithm" while "Life impacts" include "Loss of productivity," "Sleep 10 disruption," and "Safety risks."). In 2018, Meta researchers even diagrammed the interaction among 11 features that it hypothesized deliver "unpredictable, low value rewards and cues that promote frequent, 12 automatic, undesired behaviors…in a way that is not aligned with users' intentions or preferences": 13 14 15 16 17 18 19 20 21 Ex. 211 at 8566, 8578 (defining "dark pattern" features as "interactions that are deceptive, or that trick 22 you into doing something you didn't want to do" and "'addictive-like' design elements" as those "that 23 promote repeated or poorly regulated behaviors that may have a negative impact on the individual over 24 time."). Therefore, while we now proceed to a "feature by feature" analysis in keeping with the Court's 25 analytical framework, this should not obscure the reality known to Meta (and undisclosed to the public): 26 the design of Instagram and Facebook writ large present dangers to kids. 27 28 34 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 47 of 235 1 (a) Algorithm 2 Years ago, Facebook presented its users with a "News Feed" of posts from only their "friends," 3 sorted in reverse chronological order. This is not how Meta's platforms have worked for well over a 4 decade. Instead, Meta uses sophisticated machine-learning algorithms to determine what to show users 5 and in what order. And that is no longer limited to posts from a user's friends—instead, it includes 6 "unconnected" posts from other users across the platforms. Meta's change to a "ranked feed" led users "to 7 spend 40% more time in the app each day," Ex. 212 at 8565, and has been an important revenue driver for 8 the company. Ex. 219 at slide 5 (ranking "increas[es] the number of sessions and time spent when 9 compared to chronological feed"); Ex. 217 at 9529 ("The ranking team has consistently driven +0.1% 10 DAU per half for a year now!! It drives revenue! I've never seen anything like it!"). This success for 11 Facebook led Meta to begin "building a ranked Feed" for Instagram in 2015. Id. The results were similar. 12 Ex. 213 at slide 24 ("Introducing recommendations in IG Feed[.] Early tests show significant increases in 13 consumption and time spent."). Today, both platforms use algorithmic ranking across a variety of 14 "surfaces," including Facebook Feed, Facebook Notifications, Instagram Reels, Instagram Feed, 15 Instagram Explore, Instagram Stories, and Instagram Notifications. Ex. 201 at 21. 16 Meta's ranking algorithms consider various kinds of user behaviors it calls "input signals." Ex. 17 201 at 15-16. These range from extremely narrow measurements ("How many users have clicked on the 18 post in Instagram Explore to view it in full screen in the past hour," Id. at 126) to extremely broad ones 19 ("How much time the user has spent in Instagram Explore in the past 84 days," Id. at 117). Input signals 20 are fed into a complex AI model that predicts how a user will respond to a given piece of content. Id. 201 21 at 15-16. These predictions include, for example, "How likely the user is to reshare" the content, or "How 22 likely the user is to watch more than 95%" of it. Id. at 117, 121. All of these predictions are weighted and 23 summed (in what Meta calls its "value model"), and the post with the highest score gets displayed to the 24 user. Ex. 985 ¶¶ 77-78. The algorithm then runs through this exercise for the next piece of content, and 25 the next, and the next. Each video or photo is evaluated individually; their overall sequence is not 26 considered. Ex. 14 at 309:21-310:18. 27 As this discussion demonstrates, Meta's ranking algorithms do not call balls and strikes on what 28 content is "good" or "bad"—those determinations are left to separate machine learning models that 35 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 48 of 235 1 implement Meta's content moderation policies. Ex. 14 at 319:1-8 (Meta expert Birnbaum: content 2 moderation and content recommendation "are separate sets of programs"); Ex. 220 at 155:3-13 (Meta 3 expert Ferrara: "The two systems are distinct"); Ex. 986 ¶¶ 43-48 (Districts' expert Narayanan); Ex. 13 at 4 120:17-121:15 ("Facebook designs News Feed by considering how best to maximize its topline metric, 5 not the particular content to show individuals."). Put another way, Meta's ranking algorithms "respond 6 solely to how users act online." Moody v. NetChoice, LLC, 603 U.S. 707, 736 n.5 (2024).4 7 In any event, how Meta's ranking algorithms work is ultimately far less important than what they 8 are designed to do. And what they are designed to do is maximize user engagement to increase revenue. 9 That much is clear from Meta's documents. Ex. 215 at 8644 ("I feel like i'm reading totally different 10 litigation and legislation than our org leaders… we literally are optimizing teens feeds for engagement."); 11 Ex. 218 at 6686 ("the algorithm is absolutely tuned to maximize engagement in a maximally empirical, 12 principle-less way. I am not sure that the youth legal people realize this and that it applies to teens just as 13 much as gen pop. So the things contained in the lawsuits are somewhat hard to refute"); Ex. 216 at 6929 14 ("Teens are still developing, novelty-seeking and impulsive . . . Cool, so we probably shouldn't have 15 ranking primarily driven by p(eyeballtime) and p(engagement), which is how it currently works."); Ex. 16 214 at 2552 ("core algo is…literally just mathematically optimizing for [impressions] and sessions for 17 teens as with everyone else."), 2558 ("IG ranking is already different for teenagers…and FB ranking is 18 different for YA [Young Adults]…in both cases it's to squeeze even more sessions out of them."). Even 19 Meta's own algorithm expert has conceded this fundamental premise. Ex. 14 at 220:13-16 (Q: "So you're 20 not disputing that Meta's ranking systems employ user engagement objectives, are you?" A: "No."). 21 Meta has not warned the Districts, or anyone else, that its engagement-optimized ranking 22 algorithms can harm young users. Ex. 985 ¶¶ 245-253. But it knows this is the case, in several respects. 23 24 4 25 The Supreme Court has declined to address whether such behavior is protected by the First Amendment. Id.; see also NetChoice, LLC v. Bonta, 2025 WL 2600007, at *13 (9th Cir. Sept. 9, 2025) ("some 26 personalized recommendation algorithms may be expressive, while others are not, and that inquiry is fact intensive. Moody made that much clear."). The Districts expressly preserve for appeal their arguments 27 that neither the First Amendment nor Section 230 protect the operations of Defendants' ranking algorithms 28 (as distinct from their content moderation algorithms). 36 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 49 of 235 1 First, addiction is an inherent byproduct of algorithms designed to predict behavior to increase 2 usage. Ex. 239A at 1761 (FB "optimizes for time spent … it just keeps people coming back even when it 3 stops being good for them."); Ex. 13 at 108:21-24 ("advanced forms of AI that Facebook were using at 4 that point to optimize for engagement" presented "risk that that might supercharge problems of 5 addiction"), 67:12-17 ("given that the whole News Feed system is aimed at engagement, that's a set of 6 repeated patterns of behavior – clicking, liking, sharing, and so on – the best way to increase engagement 7 is addiction in some way"); see also Ex. 2A at 137:17-138:18 ("the design of the algorithm to promote 8 sustained engagement … can play a role in drawing viewers into the experience for longer periods of time 9 than they necessarily would want for themselves"); Ex. 230 at 0003 ("If algorithms favor content or 10 functionality that encourages people to spend more time on Facebook, then it's possible that this will by 11 its nature tap into addictive mechanisms" and "make addictive/compulsive usage more severe"). 12 Second, addiction is a risk created by the algorithms' unpredictable delivery of rewards. Ex. 14 at 13 134:15-18 & Ex. 223 (Meta's algorithm expert acknowledging ten reasons for variability in user feeds); 14 Ex. 108 at 5593-94 ("Maximizing for time-spent incentivizes bad ranking…we would be better off by 15 adding some randomness or filler, so that the value is stretched out…I'd be surprised if it doesn't 16 happen."). Meta understands that the variable delivery of rewards by its algorithms can lead to so-called 17 "problematic use." Ex. 107 at 0201 ("triggers contributing to PU [problematic use] habits" include 18 "Mystery of the algorithm - Uncertainty over if they will see posts from those they want; if they can find 19 a post again later."); Ex. 225 at 3133 ("[w]hen rewards are unpredictable, dopamine release is increased"); 20 Ex. 228 at 0075 ("When reward can't be reached reliably, people are incentivized to open the app more 21 frequently – but their usage becomes habitual instead of meaningful."); Ex. 139 at 0127 ("Intermittent 22 rewards are most effective (think slot machines) reinforcing behaviors that become especially hard to 23 extinguish—even when they provide little reward"); Ex. 225 at 3135 ("[i]t may be a problem if Facebook 24 seems rewarding based on the principle of unpredictability, while the inherent value of the reward is 25 lacking."); Ex. 74 at 2152-53 ("people are binging on IG so much they can't feel reward anymore"). 26 Meta knows (and has not warned) that these problems are particularly salient for its youngest users, 27 as "[a]dolescence and early adulthood are characterized by unique sensitivities to rewarding stimuli, which 28 manifest at the behavioral and neural level." Ex. 227 at 8021. Its "own product foundation research has 37 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 50 of 235 1 shown" that, because "a huge driver for teen behavior is the prospect of reward" and "[t]een brains are 2 much more sensitive to dopamine," youth "have a much harder time stopping even though they want to." 3 Ex. 224 at 0680; id. at 0676 ("IG has a pretty good hold on the serendipitous aspect of discovery…every 4 time one of our teen users finds something unexpected their brains deliver them a dopamine hit. Teens are 5 insatiable when it comes to 'feel good' dopamine effects."); Ex. 74 at 2152-53. 6 Third, because the very purpose of an engagement-optimized algorithm is to generate the most 7 engagement—by whatever means necessary—it "very frequently creates/exacerbates integrity problems." 8 Ex. 222 at 7835. It is a "a blind mathematical process with no principles." Id. The algorithm will promote 9 whatever "gets lots of engagement," irrespective of whether that is what "users like, want, and value." Ex. 10 229 at slides 5, 7; Ex. 108 at 5591-93. This means, for example, that "a comment that you hate a thing can 11 be seen as a positive signal leading content to get outsized distribution." Ex. 234 at 1252. What matters to 12 the algorithm is the fact of the comment—it does not know, or care, what the comment says. See also Ex. 13 234 at 1258 ("even Mark himself has suggested that the anger reaction is a reasonable way to express that 14 you don't like a piece of content, even as we count it as 4x more important as a less ambiguous 'like' for 15 giving you more such content"). 16 This creates additional harm for users. One former employee testified that Meta's ranking 17 algorithm "was widely understood to increase the sharing of content that provoked feelings of shame or 18 disgust or fear or pride—social emotions…that meant more user engagement on the system, and that 19 generated more revenue for the company." Ex. 13 at 122:25-123:13. "[B]y defining value in terms 20 of…clicks, likes, shares, plays – you ended up incentivizing repeated patterns of behavior on the tool that 21 were in the end undesirable, and sometimes actively harmful for the users." Id. at 81:19-24; see also Ex. 22 233 at 7777 ("While FB Feed Relevance has optimized effectively for visitation and engagement, it has 23 often increased the distribution of integrity problems–which can compound over time."); Ex. 232 at slide 24 10 ("Ranking models cause integrity regressions by increasing the prevalence of non-recommendable 25 content.") & Ex. 3 at 455:15-25 (describing this as a "known consequence" of the "choice" to "prioritiz[e] 26 recommending whatever [was] most engaging, most compelling"); Ex. 208 at slide 11 ("Meaningful 27 Interactions are highly related to harmful content. Interactions such as clicks, likes, private shares, and 28 comments are associated with sexually suggestive and hate speech content across IG."). 38 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 51 of 235 1 Meta knows this, and it knows the problem is the worst for teens. Ex. 214 at 2561 ("it seems to 2 genuinely be the case that optimizing for sessions just naturally distributes more harm to teenagers."); see 3 also Ex. 231 at slide 44 ("Bullying and Harassment (B&H), Suicide and Self Injury (SSI), and Violence 4 and Incitement (V&I)…are all >2x higher prevalence for teens than the average user"); Ex. 236 at 7851 5 ("Feed Relevance consistently distributes SSI … to teenagers and nobody has noticed."); Ex. 245 at slide 6 11 ("SSI and ED [eating disorders] have a significantly disproportionate large teen audience."). Meta 7 failed to fix this problem. See Ex. 2B at 759:14-760:12 (acknowledging that, 11 years after acquiring 8 Instagram, Meta had not addressed this problem). And it warned no one. See, e.g., Ex. 13 at 125:25- 9 126:10; Ex. 235 at 3568 (2024: "imagine going out in public and telling people we rank their kid's 10 facebook by what makes them consume the most content IDs lol" "this will be a public conversation and 11 we are not ready for it" "and then when they hear what things keep their kid consuming the most content 12 IDs … we're off to the races on a really fun conversation."). 13 Fourth, the engagement-optimized design of Meta's ranking algorithm is known to push users 14 into "rabbit holes," feeds consisting of increasingly extreme types of similar content. Ex. 985 ¶¶ 212-218. 15 Meta's would-be algorithm expert acknowledged that rabbit holing is "a plausible extrapolation from the 16 operation of the system…the content in your feed becomes more and more focused on one specific topic, 17 generally speaking." Ex. 14 at 363:21-364:8; id. at 365:20-22 (same). Importantly—and contrary to 18 Meta's Section 230 defense—he acknowledged that this phenomenon arises from the operation of Meta's 19 ranking systems, not from any piece or type of content. Id. at 366:11-16. Internally, Meta recognizes that 20 its platforms can drive users, especially teens, into dangerous rabbit holes. Ex. 242 at 1808 ("someone 21 feeling bad sees content that makes them feel bad, they engage with it, and then their IG is flooded with 22 it"); Ex. 243 at 7939 ("if you start following borderline accounts or interacting with such content, our 23 recommendations algorithms will start pushing you down a rabbit hole of more egregious content."); Ex. 24 241 at 3437 ("well-being challenges: … our ranking algorithms taking [users] into negative spirals & 25 feedback loops that are hard to exit from."); Ex. 240 at 2551 ("people who are suffering from depression 26 and self-harm go down IG rabbit holes, and explore functionality compounds this issue."); see also Ex. 27 247 at 0818 (Meta employee complaining, "I once got enticed by some snake video in watch, now all my 28 recommendations occasionally have snake videos show up. It's terrible"); Ex. 256 at 24:15-25:8 ("once 39 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 52 of 235 1 the young people start to self-injure, they tell us that their feed…on Instagram suddenly gets full of this 2 content."); Ex. 3 at 358:3-359:18. Meta also knows that its algorithm can drive adult predators into rabbit 3 holes containing child pornography. See Ex. 2A at 282:15-23 & Ex. 325 at 0324 (testifying that 4 investigators at Meta set up an Instagram test account that, within less than 24 hours, was filled with minor 5 sexualization content and recommendations for child exploitation imagery). 6 Fifth, and relatedly, the structure of Meta's ranking algorithm can trap users in "filter bubbles" 7 that validate and reinforce a single point of view. As with rabbit holes, Meta's algorithm expert 8 acknowledged that filter bubbles are "a legitimate concern" that are "a kind of reasonable extrapolation 9 from the way content recommendation systems work." Ex. 14 at 353:13-354:15. He described filter 10 bubbles as situations where "you're not really able to see a bigger picture, that you're kind of trapped in 11 a certain sense and you keep being reinforced for your beliefs and you don't see things that might 12 contradict those beliefs." Id. Again, he acknowledged this phenomenon arises from the operation of 13 Meta's ranking systems, not from any piece or type of content. Id. at 366:18-22. Meta recognizes that its 14 platforms can reinforce, validate, and amplify negative beliefs, including negative beliefs about oneself. 15 Ex. 246 at 7497 (internal comment liked 40 times: "we are far more likely to show people only content 16 they will agree with (in the pursuit of metrics), creating an echo chamber. We are…biased towards metrics 17 which ensures people's beliefs, whatever they may be, are validated and reinforced."); Ex. 248 at 5292 18 ("Preference Amplification is the process by which a user 'rabbit-holes' into a certain type of content. The 19 process may or may not be driven by intentful user actions. PA can happen for any content"); Ex. 177 at 20 slide 17 ("After connecting with accounts that include or create SSI-ED related 21 content…recommendations for suggested accounts related to SSI-ED may be shown across surfaces."). 22 Rabbit holes and filter bubbles occur because Meta's algorithms recommend content on an item- 23 by-item basis. Ex. 14 at 309:21-310:18. They do not understand when sequences of content may be 24 problematic. Multiple witnesses acknowledged that Meta's algorithms can present harm to teenagers 25 through the aggregation of photos and videos any one of which, on their own, might not violate Meta's 26 content moderation policies. Ex. 2A at 221:9-12 ("The algorithm does sometimes push content that, in 27 aggregate, could be seen as encouraging suicide"), 221:18-222:2 ("Instagram might promote content that, 28 in aggregate, could be seen as promoting self-injury," including to kids); Ex. 249 at 7437 ("IG: lack 40 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 53 of 235 1 classifiers to detect and take action on problematic aggregated content on surfaces like Explore, hashtag 2 pages, Reels."); Ex. 2A at 187:11-15 & Ex. 238 at 1685 ("when you zoom in, any one of those photos 3 probably wouldn't be triggering, but in an Explore grid with all these photos stacked against one another, 4 it's pretty overwhelming"); Ex. 2A at 191:1-20 (Explore grid for teenage girl consisting of dieting tips 5 would not violate Instagram's policies but "definitely" could create a problem). 6 Once again, Meta's erstwhile algorithm expert made these very points. He acknowledged that 7 Meta's ranking algorithms could lead "to a user seeing a lot of content…in a certain area," any of which 8 "by themselves might not be a particular problem, but taken together in the aggregate" could be 9 "potentially harmful." Ex. 14 at 330:14-333:16. Further, he admitted that this was a "potential 10 extrapolative problem of content recommendation systems," a "serious problem," and one that "content 11 moderation policies by themselves" were incapable of solving. Id. 12 As with the other problems with its ranking algorithm, Meta knew there was an issue and failed to 13 warn anyone about it. Ex. 2A at 192:16-193:15 & Ex. 238 at 1686 ("[t]here are really meaty legal 14 issues…largely by way of deceptiveness"). Nor did Meta fix the problem. Ex. 2A at 188:5-10 ("I don't 15 think we had done enough work to, like, understand or mitigate the impact of aggregated surfaces."), id. 16 at 188:21-24 ("due to just overall under-resourcing of this area, we certainly couldn't move as fast or as 17 extensively as we would have liked, given the solutions that were available to us."), id. at 195:11-196:4. 18 All of this is particularly galling given that safer, reasonable alternatives to Meta's ranking 19 algorithms do exist. See Ex. 986 ¶¶ 65-88 (Districts' expert Narayanan). For example, Meta could consider 20 the expressed preferences of its users—by simply "ask[ing] users what they think is bad," Ex. 259 at 21 2710—rather than relying so heavily on behaviors that supposedly reveal users' unstated preferences. Ex. 22 985 ¶¶ 297-301; Ex. 247 at 0818 ("we are so short term focused … most of our AI systems are using 23 behavioral user implcit [sic] signals vs Google which uses objective human ratings" / "survey based 24 metrics don't have a level playing ground with behavioral metrics like sessions."). Meta could have given 25 users the option to "fully reset their recommendations and ads," to escape from problematic rabbit holes. 26 Ex. 177 at slide 47. Meta could utilize a technique called "listwise optimization" to mitigate rabbit holing 27 and filter bubbles, Ex. 14 at 345:20-346:20, 352:22-353:2, and it understood that others in the "Industry 28 [had] successfully" done so, including "successes in Baidu, Alibaba, Kuaishou, etc." Ex. 257 at 1306. 41 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 54 of 235 1 And, of course, Meta could default users to a pure chronological feed—which Meta has long been capable 2 of doing, Ex. 281 at 7228, and which teens have long told Meta they'd prefer. Ex. 280 at 0439 (2020: 3 "Repeatedly over the past year, across topics and workstreams, we've heard teens complain bitterly about 4 Instagram's nonchronological feed…. Many say that if they could change one thing about IG, this is what 5 they would change….Although the non- chronological feed fuels increased time spent on the site, teens 6 say it makes it more difficult for them to find relevant info about their closer friends."). 7 Meta owes a duty of reasonable care to its youngest users and their caretakers, a point its own 8 employees have recognized. See Ex. 2A at 187:2-5 ("the responsibility of making sure those are 9 responsible recommendations falls to us."); Ex. 234 at 1247 ("We are Responsible for Viral Content"); 10 Ex. 248 at 5291 (acknowledging there are "Users driven to harmful content by Instagram" who are "on 11 the path to the worst experience, thanks to IG"). Yet it has failed to implement any of the safer available 12 alternatives—for the simple reason that they would reduce engagement and thus revenue. See Ex. 2B at 13 383:14-385:11 ("a significant reason" the aggregation problem wasn't solved is that it was "generally 14 pretty difficult to make safety changes that might impact growth or daily active users by any significant 15 amount"); see also Ex. 10 at 73:12-74:25 (researcher who proposed change to reverse chronological feed 16 "was laughed at the first time I brought it up" and told "we wouldn't be able to do something like that 17 because it would lower engagement"); Ex. 13 at 56:14-57:7, 121:19-122:21 (researcher on Responsible 18 AI team describing how "my team was folded and disbanded," which "demonstrated that the business 19 model was prioritized over the safety of users"). Shockingly, technical documentation reveals that Meta 20 allows changes to its ranking algorithm to regress safety more for teens than the population at large. See 21 Ex. 392 at 7539 (changes to ranking algorithm allowed to regress safety by 0.24% for Teens but only 0.1% 22 for the general population) & Ex. 14 at 372:1-5, 375:1-4, 381:6-384:12 (Meta's own expert conceding 23 "it's intuitively plausible…that this would be less protective than it should be of teens. That's an intuitively 24 plausible interpretation."). 25 Despite recognizing (and abdicating) its responsibilities to young people, Meta witnesses 26 repeatedly testified that the company "did not warn parents or kids about the specific types of risks" posed 27 by their platforms. Ex. 2A at 66:12-22. See also Ex. 168 at 144:6-15, Ex. 168B at 436:16-22 (testifying 28 that Meta did not warn parents or their children use of their platforms could lead to addiction or 42 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 55 of 235 1 problematic use); Ex. 13 at 125:18-126:10 ("I was not aware of any efforts by Facebook to warn users of 2 the potential impact of Facebook's ranking model design."); See also Ex. 997 ¶ 148 (discussing Meta's 3 failure to be transparent with users and collecting testimony). 4 (b) Infinite Scroll 5 Meta's ranking algorithm works in tandem with another feature, infinite scroll, to "remov[e] 6 stopping cues so that [a user's] brain doesn't wake up to catch up with impulses." Ex. 1070 at 41:16-18. 7 The infinite scroll is exactly as its name suggests—a never-ending feed of content through which a user 8 can thumb endlessly. This "transforms user engagement into a passive, habitual activity" that "keep[s] 9 users in a loop of reactive consumption." Ex. 987 ¶¶116-17. The feature's inventor, Aza Raskin, has 10 publicly stated that it "was one of the first products designed to not simply help a user, but to deliberately 11 keep them online for as long as possible." Ex. 252 at 2. In deposition testimony in this litigation, he 12 expressed that he "deeply regrets the unintended consequences" of the feature. Ex. 1070 at 69:4-10. 13 As early as 2017, Meta's internal documents show that the company's researchers were aware that 14 the infinite scroll could promote problematic use of its platforms. See Ex. 211 at 8566 (including infinite 15 scroll in a list of "addictive-like" features "designed to encourage a particular behavior that is not 16 necessarily in the users best interests"); Ex. 105 at 4015 (describing features like infinite scroll as being 17 "inherently at odds with well-being"); Ex. 10 at 71:14-21 ("infinite scroll allowed people to use the 18 product more than they wanted or more than they had intended to and . . . could really exacerbate people's 19 use of or problematic use of the products."); Ex. 225 at 3116 ("We should reduce cases where rewards are 20 unpredictable or lacking in value, and reduce unintentional behavior, for example: …continuous content 21 without breaks."). Despite concerns that "limitless content should come with cues for pauses, completion, 22 or active choice to continue," Ex. 253 at 6108, and the ability to implement features that would address 23 the harmful impact of infinite scroll, Ex. 254 at 5382-85, Meta's policy was to "give people control, but 24 not in a way that hurts [engagement] metrics." Ex. 255 at 7644. Meta never warned its users that its infinite 25 scroll feature could lead to problematic use of its platforms. Ex. 997 ¶¶ 162, 167. 26 (c) Screen Time Management 27 Meta knows that one way to solve problematic use is to simply limit how much time teens can 28 spend on their platforms. See Ex. 12 at 131:1-136:4 (discussing several product features that Meta 43 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 56 of 235 1 considered developing to curb problematic use). But, during the relevant time period of this case, it has 2 offered its users only weak, ineffectual, and sometimes outright inaccurate tools for this purpose. And it 3 has done so expressly to avoid limiting growth, engagement, and (therefore) revenue. 4 As of March 2020, Meta's only "proof point" on "tech addiction/problematic use" was a feature 5 telling users how much time they had spent on the platforms. Ex. 260 at 1780. ("there's no product work 6 we've done in the last four years that comes close."); id. at 1784 (tool was "flagship launch[]" that was 7 "intended to address problematic use"); Ex. 80 at 5380 ("Initial products were launched to give people 8 more control over their use, though efforts were defunded end of 2019"). But it turned out this feature was 9 tabulating users' time incorrectly. Ex. 260 at 1782 ("we're sharing bad metrics externally" which could 10 result in "legal liability."). Meta considered "unship[ping] this without a replacement"—but it decided 11 against doing so, concerned it otherwise "wouldn't have the credibility we now have in the social 12 comparison/mental health partner space." Id. at 1780. In short, until 2021, Meta offered only one tool to 13 limit problematic use, the tool was bad, and Meta left it up anyway for show. 14 In 2021, Meta considered launching a new control (called "Take a Break") to facilitate users 15 seeking to limit their use of Facebook and Instagram. "Take a Break" nudges users to stop using the app 16 after a 20-minute session. Meta knew this feature would be most effective if it was turned on by default 17 because (1) "most users don't know about the tools" Meta has, and (2) "are not likely to go into settings 18 to find them." Ex. 261 at 8248; see also Ex. 262 at 6152 ("[S]tats show us that most people just use 19 whatever the default setting or filter is."); Ex. 3 at 199:8-12 (if safety features are "not searchable and 20 they're not findable, it certainly limits their ability to help people."); Ex. 261 at 8256 (July 2019: "research 21 shows that the tools we have are hard to find."); Ex. 987 ¶¶ 111-112 (collecting sources). 22 But the team responsible for "Take a Break" was unable to even test it as an "opt-out" default, 23 because leadership viewed it as a "one-way door"— "features that once [launched], we basically cannot 24 disable" because "it looks really bad." Ex. 12 at 104:6-105:23; Ex. 263 at 9807 ("Team was unable to test 25 Take a Break opt-out default for new teens due to 'one-way door' principle."). Meta did not want to even 26 crack this door open because of concerns that a default-on Take a Break would cause "regressions" 27 (negative impacts) to "ecosystem metrics" (growth, engagement, and revenue). Ex. 264 at 7198; Ex. 265 28 at 1173 ("[W]e will: disregard [Take A Break] variants with material ecosystem regressions"). 44 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 57 of 235 1 Meta launched "Take a Break" as an "opt-in" setting—meaning, users would have to know about 2 the feature, then exercise the time and discipline to navigate the platforms' complex settings menu to turn 3 it on. This was worse for user safety and worse for reducing problematic use. Ex. 12 at 164:15-24. But it 4 was better for Meta's business metrics. See Ex. 266 at 45-46 (determining that opt-in version of Take A 5 Break would have no "ecosystem impact" on "IG Time Spent," "IG Revenue," and "Stories Impressions," 6 among other metrics). Meta knew that very few users would ever adopt Take a Break. Ex. 267 at 3662 7 (predicting that more than 99% of teen users were projected to not use Take a Break) and Ex. 113B at 8 441:21-442:2 (acknowledging same). And in fact this turned out to be the case. Ex. 982 & Ex. 270 at 3 9 (showing adoption rate for Take a Break of 0.45%). Nonetheless, Meta has publicly (and falsely) touted 10 the effectiveness of this tool. Compare e.g., with Ex. 268 at 3 (announcing the launch: "We're encouraged 11 to see that teens are using Take a Break. Early test results show that once teens set the reminders more 12 than 90 percent of them keep them on."). 13 (d) Barriers to Exit 14 Because Meta failed to implement effective tools to help users curb addictive and/or problematic 15 use of their platforms, users often attempted to self-remediate by deleting their accounts entirely. See Ex. 16 107 at 0204 (describing how addicted users would take "extreme measures" by "deleting the [Facebook] 17 app or deactivating [their account] to curb use."). But Meta intentionally designed its account deletion 18 flow to be "fairly aggressive as we try to stop people from leaving," including telling users that their 19 friends "will no longer be able to contact you through the site" and "requir[ing] users to choose a reason 20 for why they're leaving." Ex. 104 at 0087; see also Ex. 1070 at 37:7-20 (describing Facebook "come-back 21 emails that include the faces of all of your friends" which "harness[] the vulnerability [of] social 22 reciprocity and social loss"); Ex. 987 ¶ 92 (depicting the thirteen screen deactivation flow for Instagram 23 ending with a 30-day waiting period). This designs traps users in a proverbial "Hotel California"—where 24 they can try to check out any time they like but few ever leave. See id. ¶93 ("This friction…create[es] 25 barriers, wherein common actions that are potentially undesirable to the platform—such as deletion—are 26 made unduly complex to suppress a user from achieving their initial goal."). Meta does not warn its users 27 or the public that it designs its systems to discourage account deactivation and deletion. 28 45 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 58 of 235 1 (e) Notifications 2 Notifications from Meta's platforms are another known driver of social media addiction and 3 problematic use for teens. Ex. 107 at 0201 ("notifications" among the "10+ triggers contributing to 4 [problematic use] habits"); Ex. 105 at 4015 (notifications among the features designed to "provide a 5 dopamine rush"); Ex. 300 at 3213 ("21% of teen WAU [weekly active users] say notifications make it 6 harder for them to manage the amount of time they spent on the app."); Ex. 665 at 0086 ("notifications 7 can sometimes make it difficult to do our best work, might dimmish our productivity, and/or could 8 contribute to feeling of distress or being overwhelmed."). 9 Meta is aware of the well-settled science that, "when social media use displaces sleep in 10 adolescents (via nighttime social media use), it is negatively correlated to indicators of mental health." 11 Ex. 301 at 0423. "[P]reteens who get fewer than the recommended hours of sleep… were more likely to 12 have brain networks that were less efficient, less able to adapt and rewire, and less resilient to stress." Ex. 13 302 at 1532; see also Ex. 1069 at 104:20-105:7 (agreeing that consistent sleep is beneficial to adolescent 14 brain development); Ex. 981 ¶¶ 307-310 (collecting sources). 15 As such, it is jarring that Meta understood (though it did not disclose) that late-night notifications 16 are a significant driver of problematic usage for teens. Its internal research reveals that "43.3% of [weekly 17 active user] teens have at least 1 late night session"—meaning a session between 12am and 4am—and 18 that "4.6% of [weekly active user] teens have 1 or more late sessions every night." Ex. 116 at slide 10. 19 Further, "18% of [daily active users] start a session from a notification that was sent and clicked within 20 12 am and 4 am." Id. at slide 39. For teens who spend 25 minutes or more on Instagram every late night, 21 over a third of that usage results from "a notification delivered within 5 minutes of starting a session on 22 the app." Ex. 305 at 11. Clearly, Meta has a fair share of responsibility for the overall problem that "1/5 23 teens report waking up and checking social media in the middle of the night." Ex. 105 at 4007. 24 Since 2017, Meta's safety product teams have proposed launching products—such as "Off-Mode," 25 (Ex. 105 at 4009), "Together Mode," "Face-to-Face mode," "DND [Do Not Disturb]," and "Quiet Mode" 26 (Ex. 306 at 9) in response to feedback from young users who expressed that they "didn't want notifications 27 to appear during certain periods of time, such as during school… [or] while sleeping." Ex. 307 at 6249; 28 Ex. 306 at 5 ("[teens] would unanimously use a feature that limited notifications to certain times of the 46 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 59 of 235 1 day."). But Meta has consistently "shelved" these safety features, Ex. 306 at 9, or watered them down to 2 the point of being useless. According to one former employee who helped developed "Quiet Mode," Meta 3 was concerned that this feature would negatively impact metrics related to growth and usage, resulting in 4 a version less protective of teen health. Ex. 12 at 142:19-143:2. Indeed, when this feature was launched in 5 2023 (six years after it was first proposed, Ex. 306 at 9), it was buried as an "opt-in" setting and used only 6 by 0.26% of Instagram users. Ex. 308 at 3. 7 Meta has not wanted to limit notifications because notifications drive revenue. See Ex. 312 at 0649, 8 0666 (reporting that "50% of Android" and "⅓ of iPhone" daily active persons open Facebook "via 9 clicking on a push notif" and that that "these sessions enjoy [a] disproportionately higher rate of 10 engagement") (emphasis in original). Internal documents demonstrate that Meta "limit[s] the amount of 11 information in notifications because we want people to come onto the site," Ex. 139 at 0117, and 12 "optimize[s] when notifications are sent" to "improve engagement by sending notifications during the 13 right time of the day." Ex. 1178 at 8981. Specifically, Meta's notifications research has determined that 14 noon—or the middle of the school day—and 8-9 PM are "ideal times" to send push notifications to users' 15 phones. Ex. 310 at 2378; see also supra Ex. 54 at 9227 (discussing "school blasts" that sent targeted push 16 notifications to students within specific schools). Meta has even discussed the importance of using 17 notifications to make their platforms "an urgent app for US (Girl) Teens" to "bring [them] back to the 18 app[s]." Ex. 309 at 4381-82. 19 All of these decisions have been intentional and driven by "a very clear and strong message from 20 Mark [Zuckerberg] that DAP (and specifically US DAP) is extremely important" and "a bigger concern 21 [] right now than user experience." Ex. 311 at 7391-92. At no point has Meta informed its users, their 22 caretakers, Districts, or the public at large that it intentionally engineers when notifications are sent, and 23 what information they contain, to boost engagement on their platforms. 24 (f) Social Validation Metrics 25 Facebook was one of the first social media platforms to introduce the "like" button, a lightweight 26 signal that allows users to express approval of someone else's post. Ex. 355 at 0962 ("2009 FB launches 27 the like button"). Meta has known since at least 2015 that this feature can lead users to develop 28 expectations of social validation and, by doing so, increase site visits and user engagement. See Ex. 356 47 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 60 of 235 1 at 5920 (Meta research indicating that "[s]elf-motivated site visits increased after contribution . . . which 2 is consistent with the theory about feedback expectations."); Ex. 359 at slide 28 (stating that "Facebook's 3 rewards are easily accessible" and that "expectation of quick responses and availability e.g., 4 updates/counts of Likes" contributed to problematic use). Meta also knows that these dynamics are 5 particularly salient for teen users. Ex. 357 at 1216 (likes and similar signals "encourage teens to continue 6 engaging and coming back to the app" because "[a]pproval and acceptance are huge rewards" for teens). 7 In 2016, Meta expanded the ability to "like" to numerous other emoji-based reactions—a heart, a laughing 8 face, an angry face, and so on. Ex. 391 9 Meta knows that the visibility and quantifiability of social signals can impact the brain. Indeed, 10 two former Meta executives went public acknowledging this very phenomenon in 2017. See supra 11 § III.A.2.a.4; Ex. 137 (Palihapitaya: users "get rewarded in these short-term signals, hearts, likes, thumbs 12 up" which leaves them "vacant and empty" and "forces [them] into this vicious cycle" because they "need 13 it back"); Ex. 134 (Parker: "we need to sort of give you a little dopamine hit every once in a while because 14 someone liked or commented on a photo or post… that's going to get you to contribute more content…it's 15 a social validation feedback loop"). A robust body of literature (including Meta's own studies) supports 16 these observations, showing that likes lead to increased dopamine activity. See Ex. 359 at 24 ("Instagram 17 pictures with more vs. fewer 'Likes' activate the ventral striatum (VS)."). 18 Troublingly, Meta's internal research also demonstrated that, by fueling users' "need for social 19 validation," Ex. 98 at 25, and given the "visual nature of [Instagram]," Ex. 131 at 6265, visible and 20 quantifiable reactions can lead users to develop negative feelings of social comparison. See, e.g., Ex. 381 21 at 6913 ("like counts are a key driver of social comparison"); Ex. 382 at 0234 (we "can be pretty confident 22 of a causal link between Like counts and social comparison."); see also Ex. 981 ¶¶ 130-131, 147, 149-157 23 (same); Ex. 4 at 54:20-55:10. Further, Meta understood that "social comparison is a key driver of well- 24 being," Ex. 381 at 6913, and that it's linked to "multiple negative well-being outcomes (e.g., increased 25 loneliness, worse body image, and negative mood or affect)." Ex. 383 at 7852; see Ex. 382 at 0235 26 ("Seeing more posts with high Like counts (1K+) was associated with feeling worse: more negative social 27 comparison"); Ex. 127 at 42 ("at least some of this association is causal."). 28 48 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 61 of 235 1 This combination—the need for social validation and the quantifiability of popularity through 2 likes—creates a perfect storm that leads many teens, and especially teen girls, to experience negative 3 social comparison on Meta's platforms. Ex. 384 at 8545 (concluding that "negative social comparison… 4 is a significant problem on Instagram" experienced by "68% of teen girls"); Ex. 131 at 6265 (problem is 5 "historically high and growing—particularly among teens."); Ex. 1175 at 4450 (acknowledging that social 6 comparison is "a big problem that IG core mechanics make worse"); Ex. 127 at 8 (reporting "51% of 7 people experience social comparison on IG" and "33% of people have been feeling worse about 8 themselves on IG for 'several months to a year'") (emphasis in original). One researcher described outside 9 studies as "consistently point[ing] to appearance comparison as a key step towards internalization of the 10 'thin ideal' and other beauty ideals which often lead to body image issues and for some eating disorders/ 11 suicide/ self-injury." Ex. 131 at 6265. Another described "negative appearance comparison and body 12 dissatisfaction" as "early risk indicators that a teen is starting to struggle" with an eating disorder. Ex. 177 13 at slide 20; Ex. 119 at 1121 (relative to other platforms, Instagram "seen as having the highest impact" on 14 "Body, Appearance Comparison"); Ex. 19 at 284:16-287:7 (teen girls told Meta that Instagram was 15 "feeding the spiral" of negative appearance comparison, exacerbating low self-esteem and body image 16 problems); Ex. 360 at 6269. 17 In response to these findings, Adam Mosseri announced in 2019 that Meta would be launching a 18 product feature that "hid" like counts on users' posts, to help limit social comparison on Instagram. See 19 Ex. 6 at 618:14-620:6 (discussing hiding likes as a way to "depressurize the experience" on Instagram); 20 Ex. 385 ("Instagram will test hiding 'likes' in the US starting next week."). Meta researchers determined 21 that this initiative, codenamed "Project Daisy," would make users "significantly less likely to feel worse 22 about themselves." Ex. 1176 at 3256; Ex. 102 at 445:8-18; Ex. 357 at 1214 (Daisy "had a statistically 23 significant impact on reducing the frequency of 'like' comparison for both teens and non-teens."). Even 24 Mark Zuckerberg agreed that "there's something nice about not quantifying [likes]" because it "feels like 25 people are just interacting rather than playing a game." Ex. 386 at 2022-23. Nonetheless, Zuckerberg 26 remained "unconvinced rolling [Daisy] out completely…would be positive." Id. 27 After a series of tests in early 2020, Meta ultimately elected to walk back Daisy because, despite 28 improving users' well-being, it was "pretty negative to FB metrics." Ex. 387 at 6684; Ex. 388 at 5702 49 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 62 of 235 1 (showing Daisy created an "ad revenue impact of minus 1% . . . and a [Time Spent] impact of negative 2 0.29%"); see also Ex. 251 at 1789 ("Instagram will need stronger evidence than it had for the February 3 Mark review that Daisy is an important feature"). As a result, the feature was shelved until 2021, when it 4 was added as an "opt-in" feature buried in settings. Ex. 389; Ex. 367A 164:19-20 (Daisy "was not a 5 default"). Internal documents demonstrate that only 0.72% of users opted for these controls after they 6 were launched (which was somehow "lower than [Meta's] estimates."). Ex. 388 at 5703. Between January 7 2023 and March 2025, only 3.48% of youth Instagram users turned on these settings. Ex. 390 at 3. 8 Meta's failure to hide likes by default is egregious given that it was widely regarded as "one of the 9 clearest things (supported by research) that we can do to positively impact social comparison and well- 10 being on IG." Ex. 380 at 1392; see also id. ("Daisy is such a rare case where a product intervention can 11 improve well-being for almost everyone that uses our products."); Ex. 393B at 339:14-19 ("I do think 12 Daisy is one of the cases where we have done a lot of research to show that there are benefits to launching 13 it."); Ex. 394B at 431:2-15 (advocating for launch of Daisy because Meta had "concrete research" showing 14 it improved social comparison and that "the positives of that outweighed the negatives"); Ex. 395 at 80:5- 15 81:15 (agreeing that social comparison is the most important driver of well-being and that hiding like 16 counts was the best way for Meta to reduce it). 17 Internally, members of Meta's safety teams lamented leadership's decision and complained that 18 "[t]he bar for well-being (and social comparison) wins has been set too high to be achievable." Ex. 396 at 19 1460; Ex. 380 at 1392 ("[I]f we can't ship Daisy given the external comms around it and the supporting 20 research, then I am doubtful that we can ship any broad product changes with the purpose of improving 21 user well-being."). In its weakened form, the teams felt that Daisy did not meaningfully improve the well- 22 being of Instagram's users. Ex. 2A at 168:9-12 ("I think that the version of Project Daisy that launched 23 did not address the core issue of negative social comparison"); Ex. 380 at 1391 ("it will be difficult to 24 make a much stronger case [for Project Daisy]"). By contrast, the growth team was unrepentant. See Ex. 25 397 at 7912 ("It's a social comparison app, fucking get used to it"). 26 Meta has not informed the public of its decision to roll back the Daisy controls nor has it ever 27 informed them that likes and other social validation metrics on its platforms contribute to problematic use, 28 50 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 63 of 235 1 negative social comparison, eating disorders, depression, and/or anxiety. See Ex. 997 ¶¶ 172, 175, 180, 2 182-83, 188, 191-92, 197, 201-02. 3 (g) Beauty Filters 4 As discussed above, Meta had extensive knowledge that many teens experience negative 5 appearance comparison and associated mental health risks from use of Instagram. See supra at 6 § III.A.2.a.4. Despite understanding these risks, Meta failed to address the visible quantification of "likes" 7 on its platforms. But it did worse. It also began to add so-called "beautification" filters (aka "cosmetic 8 surgery effects") to Instagram. Ex. 371 at 9481; Ex. 361 at 6561 (Kevin Systrom to Mark Zuckerberg, 9 2017: "We're building skin smoothing and face transforms in-house at IG"). These filters remove 10 blemishes and change the shape of users' faces "in a way that must be achieved by plastic surgery." Ex. 11 371 at 9481. Meta knew that "teens are some of the biggest users of these filters (and teen girls in 12 particular) on IG." Ex. 274 at 7020; Ex. 362 at 8117 (filters "are overwhelmingly used by teen girls."). 13 Meta also knew that allowing these filters meant "actively encouraging young girls into body 14 dysmorphia and enabling self-view of an idealized face…that can result in serious issues." Ex. 362 at 15 8117; Ex. 129 at 1908 ("Sharing or viewing filtered selfies in Stories made people feel worse"); Ex. 983 16 ¶¶ 96-97, 101-107. A literature review conducted at Meta indicated that use of filters to manipulate selfies 17 exacerbated the "risk and maintenance of several mental health concerns, including body dissatisfaction, 18 eating disorders, and body dysmorphic disorder." Ex. 364 at 6297; Ex. 365 at 9941 ("the altering of selfies 19 appears to be connected with negative impacts on…the person posting it…in terms of mental health, body 20 dissatisfaction, and eating disorder behaviors."). Meta knew that "[t]hese extreme beauty effects can have 21 severe impacts" on "the individuals using the effects" and that "[c]hildren are particularly vulnerable," as 22 are "those with a history of mental health challenges [or] eating disorders." Ex. 366 at 135. 23 In October 2019, Meta temporarily banned cosmetic surgery filters while it consulted outside 24 experts about the issue. Ex. 371 at 9481. A "significant majority" of these experts "confirmed [Meta's] 25 hypothesis that [filters] had the potential to be very harmful, in particular to young people." Ex. 4 at 46:2- 26 47:8; Ex. 366 at 135-36 (describing consultation "with 18 experts…A large majority, including individuals 27 in the AR field, recommended prohibiting these filters, citing known impacts to body image and mental 28 health"); see also Ex. 995 ¶ 156; Ex. 982 ¶¶ 269-293. These experts further reported to Meta that 51 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 64 of 235 1 beautification filters would be particularly harmful "amongst vulnerable populations (females, youth, 2 those with a history of mental health concerns, etc.)." Ex. 371 at 9481. 3 However, Meta lifted its ban in May 2020. Mark Zuckerberg directed product teams to "allow 4 people to create and share face-altering effects on Instagram" but "not allow these effects to be discovered 5 or promoted in our Effects Gallery." Ex. 372 at 7843. One of the "Key Considerations" behind this 6 decision was "competitiveness/growth"—specifically, a concern that banning such filters would have a 7 "negative growth impact, simply because any restriction is likely to reduce engagement if people go 8 elsewhere." Ex. 371 at 9481-82; Ex. 4 at 107:14-107:21; Ex. 374 at 7297; Ex. 2 at 163:15-165:1 (filters 9 were very popular and "banning them would have felt like a competitive loss"). 10 Meta's decision drew significant criticism, even from within the company. One senior executive 11 emailed Mark Zuckerberg directly: "I respect your call on this and I'll support it, but want to just say for 12 the record that I don't think it's the right call given the risks. As a parent of two teenage girls – one of 13 whom has been hospitalized twice in part for body dysmorphia – I can tell you the pressure on them and 14 their peers coming through social media is intense with respect to body image … I was hoping we could 15 maintain a moderately protective stance here given the risk to minors." Ex. 374 at 7295; see also Ex. 5 at 16 89:20-90:5 & Ex. 375 at 3599 (different executive warning that reversal of the ban would mean "[w]e 17 would – rightly – be accused of putting growth over responsibility."); Ex. 4 at 138:10-138:24 (by lifting 18 the ban on cosmetic surgery filters, Meta "didn't prioritize or put first the well-being of teenagers and 19 kids"); Ex. 368 at 4488 ("we have no idea what the long-term effects will be for this generation that has 20 grown up comparing themselves to something that's … totally fake."). 21 Amazingly, Meta's poor decisions regarding Project Daisy and cosmetic surgery filters are not 22 even the full extent of its failure to remediate the problem of negative social comparison on Instagram. 23 Another research effort at Meta involved building an automated classifier to identify content presenting a 24 high probability of negative appearance comparison.5 If such content could be identified using AI, then 25 26 5 Respecting the Court's order concerning Section 230, the Districts do not argue that Meta should be held liable for failing to detect and remove negative appearance comparison content. They offer this 27 episode only to provide context regarding Meta's knowledge of the problem that its defective beauty 28 filters made worse, and Meta's state of mind in failing to remedy that defect or warn the public about it. 52 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 65 of 235 1 Meta could avoid algorithmically recommending it to teens and others at high risk of negative social 2 comparison. Ex. 12 at 40:18-41:16; Ex 1177 at 8073 ("we really do recommend A LOT of high-NAC 3 content to teens!"); Ex 102 at 205:18-207:12 ("for 10% of teens, 57% of the recommendations in Explore 4 are high-NAC"), id. at 218:13 ("the numbers for teen girls are higher."). Predictably—characteristically— 5 Mosseri killed the project. Ex. 12 at 44:6-10 ("I know for a fact that he did not approve it."), 47:1-13 (the 6 team was "pretty disappointed" because they "felt like they had a solution" to "a big problem" but "had 7 their hands fully tied"), 68:2-13. The data scientist and data engineer who worked on the project quit: 8 "they wanted to make a difference, and they felt like they just didn't have an opportunity to do that." Id. 9 at 48:6-8. 10 In spite of this, Meta has never warned parents or their children that beauty filters contribute to 11 problematic use, social comparison, eating disorders, depression, and/or anxiety. See Ex. 997 ¶¶ 165-67, 12 171, 174-75, 179, 182-83, 189, 191-92, 199, 201-02 13 (h) Account recommendations 14 Meta' platforms, in particular their account recommendation algorithms, have actively placed kids 15 in harms' way, facilitating what Meta internally calls "IIC"—inappropriate interactions with children. Ex. 16 5A at 370:9-11. Meta has never warned the public about the breathtaking scale of its shortcomings. 17 Let's start with some statistics, revealed in Meta's documents. Instagram has a feature called 18 "Accounts You May Follow," through which the platform recommends accounts to a user. Ex. 5A at 19 230:9-20. In 2023, this tool recommended to adult groomers "nearly 2 million minors in the last 3 20 months"—and "22% of those recommendations resulted in a follow request." Ex. 319 at 6182. The 21 situation was just as bad in the reverse. According to an internal audit in 2022, Accounts You May Follow 22 recommended 1.4 million potential IIC violators to teenage users, in a single day. Ex. 5A at 372:2-8 & 23 Ex. 320 at 2028; see also Ex. 377 at 2578 (internal 2019 analysis estimating that there are 3.5 million 24 profiles "conducting Inappropriate Interactions with Children" over Instagram DM). 25 The problem has been just as bad on Facebook. Facebook has an analogous account- 26 recommendation feature called "People You May Know." Ex. 5A at 230:9-20. Meta knew that, at least 27 through the end of 2023, "we were recommending minors to potentially suspicious adults and vice versa 28 53 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 66 of 235 1 in PYMK," Ex. 321 at 3522, and that this feature "was responsible for 80% of violating adult/minor 2 connections," Ex. 329 at 3255; see also Ex. 995 ¶ 70 (expert report of Brooke Istook). 3 Short of actually fixing the problems with its account recommendation systems, Meta at least could 4 have warned the public about the likelihood this feature would put kids in harm's way. 5 Further, Meta could have taken the simple and logical step of defaulting kids' accounts to the 6 highest privacy settings, to prevent strangers from inappropriately interacting with kids. See Ex. 995 ¶¶ 7 150, 151, 155. Meta knew this was an appropriate solution. But, as detailed below, it dragged its feet for 8 years before making the change—allowing literally billions of unwanted adult-minor interactions to 9 happen in the meantime (including those facilitated by its account-recommendation algorithms). 10 Meta's culpability can be traced back at least to the second half of 2018. In August, Guy Rosen 11 (Meta's Chief Information Security Officer) emailed Adam Mosseri (Instagram's CEO) naming the 12 problem: "someone looks me up on Instagram/Facebook because I'm a cute girl, sends me a message (or 13 even a friend request) and we start messaging. That is where all the bad stuff happens (From the "less 14 bad" tier 2 sexual harassment like dudes sending dick pics to everyone; to the tier 1 cases where they end 15 up doing horrible damage." Ex. 376 at 5891. Two months later, The Atlantic published an article titled 16 "Instagram Has a Massive Harassment Problem," featuring an image of "the Instagram brand … in a 17 Dumpster fire." Ex. 34 at 1 & Ex. 113A at 66:21-23. This article discussed the experiences of two 14- 18 year-olds who described receiving expletive-laden comments about their appearance and death threats on 19 Instagram. Ex. 34 at 2, 5. Said one: "My entire experience of high school was completely ruined by 20 Instagram harassment. It's draining, it's anxiety producing." Ex. 34 at 2; Ex. 113A at 67:20-68:10. 21 At this point, any responsible company would have jumped into the breach to address a clearly 22 significant youth safety issue. But it wasn't until close to a year later that Meta began to even study the 23 problem, let alone implement a fix. In the fall of 2019, Meta designed a research plan acknowledging that 24 its existing "privacy" setting on Instagram was, at best, a misnomer—and that "private" accounts could 25 receive a heavy volume of unwanted interactions from strangers, through mentions, tags, and most 26 importantly direct messages. Ex. 323 at 0532; Ex. 113 at 130:17-132:2, 227:13-19. Meta heard this first- 27 hand from Instagram users, through a series of hour-long interviews conducted in Los Angeles. Ex. 323 28 at 0531 & Ex. 113A at 126:17-22. This qualitative research revealed that "participants were most sensitive 54 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 67 of 235 1 to unwanted contact and interaction" on Instagram, Ex. 332 at 6902 & Ex. 113A at 139:7-13, including 2 "[u]nwanted messages that were sexual in nature," which "were particularly upsetting." Ex. 332 at 6903. 3 One 14-year-old participant reported a particularly poignant story (which easily could have been included 4 in The Atlantic article from a year earlier): 5 Fake robot accounts DM you and say watch my free sex videos and the person sends it to 50 different kinds of people. This is kind of like, hey, get a $500 Visa gift card. I did that 6 one time….I was real young. I thought I am going to get some money. It asked for my address. Plain 'ole little me, I was kind of ditzy. What it said to me after that was, we have 7 your address now, we will stalk you. I was 11. I freaked out. That's probably one of the reasons I got a private account. 8 Ex. 332 at 6904 & Ex. 113A at 141:12-22. 9 In addition to articulating the problem (again) for Meta, participants were clear on the solution. 10 "[A]lmost all participants thought that younger users and new users should be defaulted to private" 11 settings. Ex. 332 at 6901 & Ex. 113A at 135:12-20. "Teens who began with a public account, and later 12 switched to private, were very certain that their younger sibling and cousins should start with a private 13 account." Ex. 332 at 6901. Based on this feedback, the Meta researcher in charge of the study made two 14 recommendations: "default[] all teen accounts to private mode" and, "for private mode defaults, limit 15 tagging, mentioning, and group DMs to connected accounts." Ex. 113A at 147:5-148:2, 149:11-25. "This 16 is more in line with users' expectations and desires." Ex. 332 at 6907 & Ex. 113A at 149:1-9. 17 At this point, Meta could have implemented changes to ensure its safety tools lined up with user 18 expectations—and, at least, warned users that Instagram's existing private mode was less protective than 19 they expected. Meta did neither of these things. Ex. 113A at 145:25-146:25. Instead, it asked its growth 20 team to study the "growth and engagement impacts of defaulting teen accounts into a private setting." Ex. 21 7A at 206:5-9. The growth team was pessimistic ("This will likely smash engagement, DAP, MAP etc") 22 and skeptical ("What's the rationale for the push?"). Ex. 342 at 5175. So Meta put off the issue. 23 When it picked things back up, it again limited itself to exploratory studies. This time, in March 24 2020, researchers embarked on an investigation into what they called "Smart Defaults"—"a safer account 25 model for teens" that would "[p]rotect teens from unwanted DMs" and "additional interaction[s]" in order 26 "to keep them safe." Ex. 113A at 156:18-157:12; Ex. 333 at slides 1, 4. In line with the 2019 interviews 27 in L.A., this research acknowledged that privacy settings for "mentions and tags are expected" and that 28 55 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 68 of 235 1 failing to make this change would "not match users expectations." Ex. 333 at slide 4; Ex. 113A at 159:24- 2 160:8. Nonetheless, the research proposed not just the option of "default[ing] all new teens into private"— 3 which it acknowledged would "keep more people safe from unwanted DMs, tags, and mentions." Ex. 333 4 at slide 16. It also outlined the option of "upsell[ing]" teens into an "opt-in" private experience—noting 5 this would present "less risk to engagement metrics." Ex. 333 at slide 16. Indeed, the growth team had 6 been called on once again. It had determined that a true private-by-default would result in a loss of 1.5 7 million monthly active teens a year. Ex. 7A at 377:8-14; see also Ex. 334 at 5 ("Your data pretty succinctly 8 shows that taking away unwanted interactions via private default settings is likely to lead to a potentially 9 untenable problem with engagement and growth."). 10 Faced with a clear choice, Meta chose profits over safety, informing its well-being researchers in 11 April 2020 that "We will not create any new default interaction settings for private accounts." Ex. 334 at 12 1; Ex. 113A at 181:16-182:1. Worse, leadership asked these researchers to gin up a rationalization for this 13 decision—"give us a story that frames our decision that we've already made" by "fram[ing]" this as a 14 tradeoff between "safety vs value." Ex. 334 at 6. Privately, researchers were disgruntled. For them, this 15 was "pretty much [the] worst-case scenario." Ex. 335 at 6621. They knew that making teen accounts 16 private by default was necessary, because users "never actually went into the settings menu" to 17 affirmatively opt-in to safety tools. Ex. 113A at 204:10-22; see Ex. 261 at 8256 (July 2019: "Current 18 research shows that the tools we have are hard to find."); Ex. 2A at 72:14-16 ("most users…don't charge 19 their default settings until they are in a moment of crisis."); Ex. 322 at 36 (acknowledging that only 0.08% 20 of youth use Instagram's opt-in tool to limit comments and DMs from strangers). One researcher 21 grumbled: "ISnt safety the whole point of this team? [sic]" to which another responded: "The point of this 22 team is to prevent another Atlantic article about how shitty instagram is." Ex. 335 at 6621. 23 Despite privately expressing the belief that Meta should "man up and default people," id., however, 24 these researchers obliged their superiors. In a note published internally on April 3, 2020, they endorsed 25 the idea "that we keep the current default interaction settings as-is," arguing that a change to the defaults 26 that limited tags, mentions, and DMs from strangers would "prevent[] an important way that new users 27 form positive connections with people they don't know" (and offering only a passing acknowledgment 28 that such a change also "would significantly impact growth"). Ex. 1183 at 8558. A month later, one of the 56 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 69 of 235 1 two researchers who authored this note was promoted. Ex. 113A at 254:5-9. 2 Meta knew that placing teens into a default-private setting would have eliminated 5.4 million 3 unwanted interactions a day over Instagram direct message—including interactions like the harrowing 4 ones described by The Atlantic two years earlier, and by interviewees in Los Angeles a year after that. Ex. 5 113 at 236:4-10; Ex. 336 at slide 17. It could have made a simple product change to keep kids safe. Having 6 elected not to do the right thing, it at least could have warned the Districts, the public, parents, or kids of 7 this danger on Instagram. It didn't do that either. Ex. 113A at 248:24-249:11, 253:19-254:3. 8 Five months after Meta's decision, the nightmare Meta could have avoided came to pass. 9 Employees determined that inappropriate interactions between adults and minors had skyrocketed on 10 Instagram Direct Message—to fully 38x what the company was seeing on Facebook Messenger. Ex. 337 11 at 0623; Ex. 210 at 2309. Meanwhile, Meta was "automatically removing 90 times less amount of 12 inappropriate interactions with children on Instagram than on Facebook." Ex. 2A at 289:19-22 (emphasis 13 added); Ex. 337 at 0623. Meta understood their role in this problem—"[a]ctors take advantage of our tools 14 on Instagram to find and inappropriately engage with children," Ex. 337 at 0623—yet they continued to 15 say nothing to the public. Ex. 113A at 260:24-261:14. Moreover, Meta knew that the problem was being 16 compounded by the launch of Instagram Reels in August 2020. By broadcasting "viral" videos of young 17 teenagers to a wide audience of adult strangers, Reels put kids in harm's way. Ex. 3 at 55:25-56:18 ("young 18 women who were producing content on Reels were experiencing unwanted interactions as a result of their 19 content being shown to people that they did not wish to have it be shown to."), 57:25-58:9 (same). 20 Researchers who suggested "filter[ing] out minor accounts from Reels altogether" to avoid this problem 21 were ignored. Ex. 2A at 47:1-48:13 & Ex. 207 at 9534; see also Ex. 3 at 80:18-81:20, 106:17-108:6. 22 Internally, the controversy over Private by Default renewed. Once again, the growth team was 23 asked to examine "the growth and engagement knock-on effects" of rolling out such a feature, including 24 potential "friending losses" and "engagement declines." Ex. 7A at 212:2-12, 215:1-5; Ex. 343 at 8137. 25 The growth team remained pessimistic ("we'll never really mitigate enough") and it remained skeptical 26 ("This was never really about increasing safety" / "It was all for the pR wins" [sic]). Ex. 343 at 8137-38. 27 In August 2020, it determined that private by default would result in a "2.2% hit to US DAU" and a "2.2% 28 hit to global for teens"—"that is HUGE." Ex. 344 at 8209; Ex. 7A at 220:2-6; Ex. 345 (tabulating impact 57 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 70 of 235 1 on growth metrics); Ex. 346 at 1386 ("defaulting people to private on day 1 means that people just use the 2 platform a lot less"). 3 In August 2020, Meta's leadership was presented, once more, with a decision. Meta's policy, legal, 4 communications, privacy, and well-being teams all recommended that private-by-default should "Launch 5 Now," Ex. 347 at 6982, observing that this "will increase teen safety" and is "in-line with teen user 6 expectations," "parent expectations," and "regulator expectations," Ex. 347 at 6983. The growth team— 7 and only the growth team—recommended "Don't Launch (Now)," noting a likely impact of negative 8 "2.2% topline teen DAP in 5 years." Ex. 347 at 6982; see Ex. 7A at 380:6-22; Ex. 7B at 703:9-19. Meta 9 did not launch private by default in 2020. Ex. 7 at 236:2-3, 384:23-24; Ex. 352 at 3254 ("the growth impact 10 was too high"); Ex. 353 at 5759 ("scrapped due to growth concerns"); Ex. 3 at 68:25-71:7. 11 In March 2021, the discussion was revived once again. Ex. 348 at 7502 ("Guess what, they're 12 reviving private by default LOL" / "mygod"). This time—at last—leadership decided to launch a version 13 of the feature. On March 16, 2021, Meta publicly announced that it would be "restricting DMs between 14 teens and adults they don't follow." Ex. 338 at 2. Meta represented that, with this change, "when an adult 15 tries to message a teen who doesn't follow them, they receive a notification that DM'ing them isn't an 16 option." Ex. 338 at 3. Not surprisingly, however, there were some significant problems—beyond the fact 17 that Meta's year-long delay had already facilitated 1.9 billion (=5.4 million x 365) unwanted interactions 18 between adults and minors on Instagram. 19 "[C]ritically, private by default was originally supposed to be for all users;" what launched instead 20 "was for new users under the age of 16"—not existing teen users, new 17- or 18-year old users, or new 21 teen users who lied about their birthday. Ex. 2B at 748:5-11, 749:1-3, 749:21-750:4; Ex. 2A at 71:9-21; 22 accord Ex. 7 at 241:13-14. Meta knew this was a problem. In July 2021 (after the launch), it learned from 23 an internal survey that 13% of 13- to 15-years-old had received unwanted sexual advances on Instagram 24 in the past seven days, overwhelmingly from strangers. Ex. 145 at 5033, 49; Ex. 168 at 307:10-25, 319:8- 25 23 ("they are experiencing these issues…facilitated by the product design of Instagram"). 26 Further, what launched wasn't actually a "true default setting." Ex. 2A at 74:13-75:2. As 27 Instagram's former Head of Safety and Well-Being testified, "if you said to somebody, oh, this is an 28 account that is private by default, I think an assumption would be, I sign up, I'm a teenager on the platform, 58 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 71 of 235 1 my account is private. I will go into settings and change that to be a public account if I want to." Id. at 2 73:8-13. Instead, what Meta launched was a "default preselected option on a screen," id. at 73:1-2— 3 meaning, new users who self-identified as 16-or-under would "get a screen asking me if want to have a 4 public account or a private account, and the private option is selected." Id. at 73:14-74:1. Meta knew this 5 would result in fewer private accounts and knew this would be less protective of teen safety, but did it 6 anyway because of growth concerns. Id. at 74:20-75:16. As one employee groused internally: "[this] is 7 about looking good to regulators so that they don't block our under 13 year old IG version we are working 8 on. Thats it. It has a terrible impact on teen engagement and retention and no detectable benefit on integrity 9 metrics." Ex. 349 at 6694. 10 Moreover, while Meta's public announcement indicated that adults would not be able to DM 11 minors with a private account, see Ex. 338 at 3, this belied the spotty execution of this tool. That was cast 12 into stark relief a year later when a "sitewide emergency event" in February 2022 revealed that teens were 13 still "receiving DM requests from unconnected adults, breaking [the] public commitment that we made." 14 Ex. 339 at 2959; Ex. 113A at 271:17-23. Indeed, fully "50% of message requests" to teen users were still 15 coming from adults. Ex. 339 at 2960; Ex. 35 at slide 15. The problem appears to have had numerous 16 causes. First, "the logic to block DM requests from adults to minors" did not include users whose age 17 Meta couldn't predict. Id.; Ex. 113A at 280:4-281:14. Meta knew this was dangerous because it knew 18 groomers regularly claim to be minors—and "[o]ur age models are totally blind to this extraordinarily 19 obvious exploit." Ex. 340 at 1116. Second, Instagram "still allow[ed]" teens to receive DM requests from 20 "senders with no stated age." Ex. 35 at slide 14. Third, it allowed such DM requests from "senders outside 21 the US who state they are teens, but may not be teens." Id. 22 Taken together, "unwanted DM requests" remained a serious risk issue for teens fully a year after 23 Meta's "private by default" announcement, with Meta conceding (internally, never externally) that "there 24 have been many gaps in fulfilling our promise." Ex. 35 at slide 15. Meta never corrected the 25 representations in its March 2021 blog post to say that adults could still interact with minors on Instagram. 26 Ex. 351; Ex. 113 at 290:25-291:11. 27 Meta did not apply private-by-default settings to all teen accounts (including existing users) until 28 the end of 2024, well into this litigation. Ex. 2A at 71:14-20. Given the depressing history told above, 59 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 72 of 235 1 "Instagram Teen Accounts" warrants skepticism. Ex. 341 at 1. According to Meta's highly-publicized 2 rollout, this feature places teen users by default into "the strictest messaging settings, so they can only be 3 messaged by people they follow or are already connected to." Ex. 341 at 3. Of course, this is precisely the 4 intervention that Meta's safety researchers had been advocating for (largely unsuccessfully) since the 5 middle of 2019. A skeptical reader might wonder what has changed—other than five additional years of 6 public pressure, scrutiny, and litigation. The discovery cut-off in this litigation has prevented the Districts 7 from adequately investigating Meta's internal understanding of Teen Accounts' efficacy. 8 What the Districts can comfortably say is that Meta delayed defaulting teens into private accounts 9 for years, prioritizing its own profits over preventing billions of inappropriate interactions between adults 10 and kids. See Ex. 350; Ex. 354 at 8526 ("it was a huge growth hit like shockingly so" / "we did NOT do 11 it for a very long period of time because of that" / "it's absolutely mad that it was so contentious" / "would 12 [Mosseri] want any tom dick or harry being able to see all his kids' content, follow them etc? is he fucking 13 nuts?"). Meta failed to warn parents, students, teachers, school administrators, or for that matter anyone 14 about these issues. See, e.g., Ex. 5A at 388:11-389:15; Ex. 6 at 239:5-21; Ex. 113A at 146:7-25, 260:17- 15 261:14, 289:9-25, 291:3-292:9; Ex. 168A at 329:24-331:12; Ex. 2A at 315:9-17. It abdicated its 16 responsibility. 17 (i) CSAM Reporting 18 Instagram and Facebook both allow users to report content that may be harmful. This is good in 19 theory. But it only works if the reporting mechanism is effective, and if Meta does something with the 20 reports it receives. Unfortunately, this is not the case for the most horrifying content that exists on Meta's 21 platform, child pornography (sometimes called child sexual abuse material ("CSAM") or child 22 exploitation imagery ("CEI")). Nor is it the case for repeat offenders who traffic children. 23 Meta went out of its way to stymie user reports of CSAM. As of March 2020, Meta did not have 24 a specific way for people to report CSAM on Instagram (whether or not logged in as a user). Ex. 2A at 25 274:6-20; Ex. 244 at 5735 ("We don't have an in-app reporting option for CEI that I'm aware of."). When 26 Vaishnavi Jayakumar, Instagram's then-new Head of Safety and Wellbeing, learned about this, "it was 27 very surprising to me," "[b]ecause reporting CEI is sort of baked into most platforms." Ex. 2A at 275:23- 28 276:2. Ms. Jayakumar raised this issue "multiple times," Ex. 2A at 276:7-8, but was told that it would be 60 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 73 of 235 1 too much work to build and, if Meta did so, it would be forced to review any reports that came in through 2 that channel, meaning even more work. Ex. 2A at 275:4-276:15. 3 At deposition, Ms. Jayakumar testified that this approach did not strike her as putting the safety of 4 kids first. Ex. 2A at 276:16-19. Contrary to what she was told, "[i]t is not a significant amount of product 5 work" to allow users to report CSAM and CEI within Instagram—"you essentially add an additional 6 option to the existing number of reporting…options that are out there." Ex. 2A at 277:8-12. Notably, 7 Instagram at the time already allowed users to report a number of other far less serious violations directly 8 within the ap. Ex. 2A at 276:21-277:2; see Ex. 182 at 8871 (photo of "IG User Reporting Flow" with 9 dedicated options to report "spam," "intellectual property violation," and "promotion of firearms," among 10 others, but no option to report CSAM). Without question, it would have been significant additional work 11 for Instagram to review reports of CEI—but that is precisely the work Meta should have done to keep kids 12 safe. Ex. 2A at 277:13-17. Meta not only didn't do this, it made its reporting flow intentionally 13 complicated, to reduce the reports it had to look at. See Ex. 168A at 163:16-166:4; see id. at 161:10-22 14 ("only 1 percent" of people who have bad experiences on Instagram "finished submitting feedback through 15 the reporting tool."). 16 Meta also failed to inform users that they could report CEI and CSAM sent via ephemeral message. 17 "Disappearing Mode, which is also known as Vanish Mode on Instagram today," is a feature that "allows 18 you to send messages to people and then have it disappear, you know, right after." Ex. 2B at 457:1-5. 19 Because these messages are, by their very nature, ephemeral, users are unlikely to know they could be 20 reported to Meta. Ms. Jayakumar testified that, "in order to protect minors who might be subject to 21 inappropriate messages or grooming messages, we wanted to upsell and really promote the value of 22 reporting within Disappearing Mode…to educate the user that even ephemeral messages can be reported 23 and that they have a limited window of time in which to do so." Ex. 2B at 457:9-19. However, Meta 24 expressly declined to undertake that educational effort—again, because "the team that would review the 25 reports… wouldn't have the resources" to do so. Ex. 2B at 458:2-10. 26 Disturbingly, Meta's desire to look the other way extended even to the accounts of repeat 27 predators. Ms. Jayakumar testified—and internal documentation confirms—that Meta had a "17x" strike 28 policy for accounts engaged in the "trafficking of humans for sex." Ex. 2B at 459:9-23 & Ex. 199 at 7242 61 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 74 of 235 1 ("This area is severely under-resourced given our SESTA-FOSTA liability."). Ms. Jayakumar explained: 2 "That means that you could incur 16 violations for prostitution and sexual solicitation, and upon the 17th 3 violation, your account would be suspended." Id. This, "by any measure across the industry, is a very, 4 very high strike threshold." Ex. 2B at 460:9-13. Needless to say, Meta never told parents, the public, or 5 the Districts that it doesn't delete accounts that have engaged over fifteen times in sex trafficking. Id. at 6 315:9-17. 7 (j) Age Verification 8 In September 2024 (well after these lawsuits were filed), Meta announced what it claimed to be a 9 new suite of protections on Instagram, "Instagram Teen Accounts." Ex. 271 at 1. Meta has told the public 10 that it will "automatically place teens in Teen Accounts, and teens under 16 will need a parent's permission 11 to change any of these settings to be less strict." Id. at 1. It has assured the public that Teen Accounts 12 provides "Built-In Protections for Teens [and] Peace of Mind for Parents." Id. at 2. 13 But the fundamental problem with Teen Accounts is the same problem that infects all of Meta's 14 other loudly-touted youth safety tools, such as Private by Default. Supra at § 6(h). Meta does not know 15 the ages of its users, and it cannot protect those it cannot identify. As whistleblower Jason Sattizahn 16 explained in testimony to Congress, the "Teen Accounts announcement … elides the fact Meta does not 17 have adequate data or processes to accurately track teen users"—though it "does have nearly a decade 18 worth of research showing that Meta does not know the ages of its users." Ex. 273 ¶¶ 140-146. 19 Instagram did not even require new users to provide a birthdate during account setup until 20 December 2019. Ex. 274 at 7020. It wasn't made mandatory for existing users until 2021. Ex. 195 at 7; 21 Ex. 113 at 79:23-80:2. As a result, for years, Meta did not even have self-represented age for an enormous 22 fraction of Instagram's user base. Ex. 240 at 2552 (Apr. 2020: "we also only have age for 55ish percent 23 of people"). In any case, Meta knew that simply asking users to input a birthday would be ineffectual. Ex. 24 9A at 72:8-73:12 ("pretty straightforward for people to enter a different age"). And it was. Meta is aware 25 that at least "30%-40% of teens are registered as being >18." Ex. 25 at 13. Some internal estimates put 26 this number even higher—at 60% of teens. Ex. 194A at 107:12-108:2. Further, data produced in this case 27 reveals enormous "blind spots" in Meta's understanding of the ages of its users. See, e.g., Ex. 282 (users 28 with "unknown" predicted age: 142 million (Feb 2018), 162 million (Feb 2019), 182 million (Feb 2020), 62 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 75 of 235 1 204 million (Feb 2021), 216 million (Feb 2022); id. (in Feb 2018, it was "unknown" whether 16.2 million 2 monthly active users were "teen" or "non-teen"); Ex. 283 (authenticating data). Needless to say, "Built- 3 In Protections for Teens," Ex. 271 at 2, won't work if Meta can't even identify who those users are. See 4 Ex. 2B at 476:6-17, 730:23-732:16. 5 Meta's lack of understanding of the ages of its users is problematic for an additional reason—it 6 leaves Meta unable to comply with its obligations under federal law. Meta is required to observe numerous 7 data privacy safeguards for any user under the age of 13 years old. See 15 U.S.C. § 6501 et seq. 8 ("COPPA"). In this litigation, Meta has admitted it does not observe any of these protections. Specifically, 9 "since 2012, Meta has not obtained verifiable parental consent for individuals under the age of 13 to use 10 Instagram," Ex. 184 at 9, it "has not provided parents means to review personal information for individuals 11 under the age of 13," Ex. 184 at 11, and it "has not provided parents a means" to "refuse to permit the 12 further use of personal information collected from their children under 13." Ex. 203 at 11-12. Compare 13 id. with 16 CFR §§ 312.3-5 & 16 CFR § 312.6. 14 Meta's justification for its noncompliance is that it "does not permit individuals under 13 on its 15 platforms." Ex. 190 at 4; see Ex. 191 at 1 (same). This would make sense if it were true. But it is not. In 16 fact, for nearly a decade, Meta looked the other way when it came to underage users of its platforms. 17 Multiple company witnesses acknowledged that Meta is "generally aware … that there are people who 18 are under the age of 13 who lie about their age in order to get around the controls that different apps have." 19 Ex. 11A at 425:16-426:17 (Zuckerberg); accord Ex. 286A at 110:20-111:3 (Antigone Davis); Ex. 113A 20 at 216:20-217:24 ( ); Ex. 287 at 3666 ("definitely kids this age on IG"). But Meta's knowledge 21 is not just generic. As early as 2017, Meta knew that "FB and IG currently predominantly function as 22 consumption platforms for tweens." Ex. 317 at 4804. It knew that "[t]eens typically join Instagram when 23 they are in middle school at age 11-12." Ex. 275 at 6272-6273 ("timeline of Instagram usage"). It even 24 observed that "Instagram Sells Itself: In 6th grade, age 11-12, teens get first phone and are 'invited.'" Ex. 25 314 at slide 26; see also Ex. 296 at 4832 (quoting "11yo girl": "More people my age are into Instagram."). 26 By 2018, Meta knew "there were 4M people under 13 in 2015 on IG" which "represents around 30% of 27 all 10-12 years [sic] old in the US." Ex. 192 at 3546. It knew that "45[%] of children 9-12 say they use 28 Facebook daily" while "40 percent use Instagram." Ex. 193 at 5502. 63 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 76 of 235 1 These numbers are hardly an accident. They are the foreseeable result of Meta's persistent interest 2 in capturing users as early as possible—because "Tweens and young teens are Facebook's best 3 opportunity to win future generations of people." Ex. 315 at 1198; see also Ex. 313 at 5642 ("Tweens 4 (approximately age 10 to 12) are special. People who join[] Facebook as tweens have the highest long- 5 term retention out of all age groups."); Ex. 316 at slide 4 ("We have definitively established tweens as the 6 highest retention age group in the United States."). 7 That interest has manifested in extensive research into the psychology and digital behavior of 8 "tweens." See, e.g., Ex. 297 at 8628 ("Objectives: To unearth digital behavior of 8-12s"); Ex. 294 at 1309 9 ("Tweens are herd animals, and want to find communities where they can fit in."); Ex. 296 at 4809 10 ("Tweens are already on social media."), 4821 ("The majority of 10-12 year olds have at least one social 11 media account."), 4845 (outlining research plan for social media users "approx age 10-13"). It has also 12 included exploratory missions into new products designed for "users as young as 5-10." Ex. 293 at 1556; 13 see Ex. 295 at 4410 ("Exploring playdates as a growth lever for Messenger Kids"); Ex. 298 at slide 13 14 ("5-10 year olds are our best initial U13 segment, allowing us to build[]on FB parent strength to grow 15 a portable graph for young kids who will become future teens.") (emphasis in original). 16 Internally, employees were nothing short of disgusted at Meta's express interest in the tween 17 market: "oh good, we're going after <13 year olds now? [Z]uck has been talking about that for awhile… 18 targetting [sic] 11 year olds feels like tobacco companies a couple decades ago (and today). Like we're 19 seriously saying 'we have to hook them young' here." Ex. 299 at 4290. 20 Meta has not acknowledged the widespread use of its platforms by tweens. Instead, it has denied 21 the problem, "avoid[ing] conversations that highlight <13 yo with IG accounts," Ex. 279 at 8207, and 22 declining to "ask [users] if they're under 13 because it means we'd have to delete their account," Ex. 285 23 at 3406; see also Ex. 168A at 183:25-184:4 (testifying Meta had a "don't ask, don't tell" policy when it 24 came to underage use of Instagram from 2012-2015). When pop star Jojo Siwa told Adam Mosseri she 25 had started an Instagram account at the age of 8, Mosseri's response was that he "did not want to know." 26 Ex. 292 at 15. (Mosseri subsequently copped to the Senate that this was a "missed opportunity." Id.). Meta 27 has even ventured beyond denial into outright deception, publishing "community standards pages 28 mentioning we age gate"—though, in reality, this was "not implemented in IG." Ex. 276 at 5781; see id. 64 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 77 of 235 1 ("we don't even have age on IG for a big chunk (last figure I saw as ~30%) of IG users...I wonder if the 2 authors of those web pages know that?"). 3 Meanwhile, Meta failed to provide users with good tools to report under-13-year-old users. See 4 Ex. 196 at 6516-17 ("The [under 13] reporting flow was pretty bad….It was obviously structured to deter 5 any reports."); Ex. 2A at 98:25-101:13 (reporting flow "very clunky" and resulted in "a lot fewer reports 6 than otherwise"); Ex. 197 at 8850 ("currently there is no way for users to report U-13 users on Facebook 7 in app"). And it failed to review millions of the reports it did receive—between 2020 and 2021, it 8 maintained a backlog of between 450,000 to 2.7 million suspected under-13-year-old accounts. Ex. 199 9 at 7246; Ex. 200 at 6780; Ex. 288 at 2800 ("big backlog and demand is outpacing supply"). Removal of 10 these accounts was deemed "not a priority," despite the fact that it meant "our basic COPPA compliance 11 is at risk." Ex. 198 at 6543; Ex. 2A at 90:3–4 ("there weren't staff allocated to reviewing and addressing 12 the backlog"); see Ex. 174 at 5153 ("The fact that we have age limits which are unenforced 13 (unenforceable?)…makes it difficult to claim we are doing all we can."). 14 Meta could easily implement processes to find and remove under-13-year-old users (to comply 15 with federal law) and correctly identify under-18-year-old users (to give its anemic safety tools at least 16 some shot at success). Numerous options were, and remain, available. See Ex. 171 at 6015 ("if we used 17 age classifiers we could detect under 13s and kick them off the platform."); Ex. 272 ¶ 14 (declaration of 18 Sattizahn); Ex. 989 ¶¶ 217-264 (Estes opening report). Indeed, Meta has used such tools for the benefit of 19 its advertisers. Ex. 276 at 5780 (2021: "in the past, the only requirement for age-gating on IG was by 20 advertisers, around brand safety. That's why the relevant infra and code are owned by the IG Ads team"); 21 Ex. 2 at 464:10-21 ("we sometimes…[would] try and predict the age of that account, to decide what kind 22 of ads might be most relevant to them."). Meta has also used ID verification in the European Union, to 23 prevent adults from misrepresenting themselves as teens "to get an ad-free experience." (The EU prohibits 24 ad targeting to minors.) Ex. 290 at 1446. While Meta could easily use ID verification to "identify[] if a 25 user is a teen," it did not do so because of "the potential revenue impact it can cause." Ex. 291 at 1601. 26 As one former employee observed, if Meta can conduct age assurance "for business purposes," it "should 27 also be able to do that for integrity purposes." Ex. 2B at 464:10-21. 28 65 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 78 of 235 1 But Meta's executives have stifled any efforts to implement new methods of age assurance to 2 remove underage users or protect kids. See e.g., Ex. 174 at 5145 (Bickert, 2019: "Would we really want 3 to have public statements from our expert that Instagram needs to have age gating[?]"); Ex. 198 at 6543 4 (2020: "Adam Mosseri will not support upselling age collection for existing users because he sees it as 5 intrusive data collection."); Ex. 284 at 6683 (2020: "no explicit decision on improving our existing age 6 models. The leadership decision was limited to NOT building an u13 classifier") Ex. 199 at 7245 (2020: 7 "Overall we are quite vulnerable when it comes to age management on our apps. To some degree that is 8 because leadership has decided [to] live with certain risks at least in the near term."); Ex. 272 ¶ 6 9 (describing "sudden cancellation" of age assurance research project in 2022). Meta simply will not do so 10 unless it is required. Indeed, after Arkansas passed a mandatory age verification statute, Meta built an age 11 verification tool that was "ready for launch"—but it "paused" its implementation when the law changed. 12 Ex. 289 at 1568-1569. 13 (k) Parental Controls 14 Meta's delays in launching parental supervision features on Facebook and Instagram have been 15 unacceptable. In 2009, Facebook's founding engineer sent Mark Zuckerberg and Chris Cox an email 16 proposing parental supervision tools. "I would like to see us add an opt-in feature, which would allow a 17 Facebook user (child) to designate another user (the parent) to have certain auditing rights and limiting 18 controls over the child account," he wrote. Ex. 318 at 5130. "The dynamic that this creates is to give 19 parents an opportunity to act as parents on Facebook as they would in other dimensions of their children's 20 lives … which I believe is the only scalable and effective way to address the issues of minors on 21 Facebook." In reply, Chris Cox (now Meta's Chief Product Officer) responded that this "doesn't 22 immediately strike me as high-leverage as much as nice to have" and "seems like a bunch of work for 23 things that the parent could do manually." Id. 24 Meta did not make parental controls available to Facebook users until 2023—fully 14 years later. 25 See Ex. 184 at 7 ("Meta admits that parental controls, including tools developed for parental supervision 26 of Teens, were first made available to users on Facebook in 2023."). It has offered no explanation or 27 justification for this delay. Nor has it explained why only 0.15% of Youth users are enrolled in these 28 parental supervision tools—meaning, 99.85% are not. See Ex. 168B at 581:17-582:10 (Béjar: making 66 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 79 of 235 1 parental supervision opt-in meant "that the feature would not be adopted and then as such would not be 2 effective as a safety feature"). The timing and adoption rates speak for themselves. Facebook's parental 3 supervision tools have been nonexistent or ineffective during the relevant time period of this litigation. 4 The situation is no better with respect to Instagram. Meta only launched parental controls on 5 Instagram in 2022, ten years after it acquired the platform. Ex. 184 at 6 ("Meta admits that parental 6 controls, including tools developed for parental supervision of Teens, were first made available to users 7 on Instagram in 2022."); Ex. 2B at 736:11-739:20. Further, "[t]he initial version of parental supervision 8 required both parties to agree to supervision, so the teen or the parent could disable supervision." Id. at 9 737:21-24. Not surprisingly, adoption rates were as dismal as they were on Facebook—with only 0.38% 10 of youth enrolling in supervision on Instagram from November 2023 through March 2025. Ex. 322 at 39; 11 Ex. 308 at 3. Meta ultimately took an "about-turn from th[is] position" with the launch of Teen Accounts, 12 id. at 738:16-21—and, without a hint of irony, even sent Instagram's CEO on the Today show to tout the 13 importance of safe defaults. Ex. 328 (Adam Mosseri, Oct. 14, 2025: "I've got three kids, I know parents 14 are busy, which is why it's so important that the defaults are safe."). But the fact remains that it took 12 15 years after acquiring Instagram for Meta to adopt parental supervision tools that a teen couldn't just disable 16 on their own. Id. at 739:15-25. And even today, the parental supervision tools on Instagram Teen Accounts 17 are only as good as the age verification behind them. Those tools remain broken. See supra § (6)(j). 18 At this point in the brief, the reader will hardly be surprised to learn why Meta delayed for as long 19 as it did in releasing parental supervision tools. Parental involvement in teen accounts was seen as a threat 20 to growth—limiting teens' content production, sharing, and network development. Ex. 327 at 1975 ("We 21 should be thinking about how parents being on Instagram might effect [sic] graph management and teen 22 engagement over time."); Ex. 298 at slide 14 ("Adults are Toxic: Unless we allow teens to share separately 23 from adults, the rest of the world will inevitably follow the decline of teen engagement in the US and 24 EU."). More succinctly: "Parents are a killjoy." Ex. 324 at 8318. 25 Needless to say, this is no excuse for depriving parents of the tools Meta knew they wanted, 26 needed, and were entitled to in order to keep their kids safe. See, e.g., Ex. 98 at 9218 ("parents can't 27 understand and don't know how to help" because "social media has fundamentally changed the landscape 28 of adolescence"); Ex. 177 at slide 11 (parents "do not have the ability to take direct action to protect their 67 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 80 of 235 1 teens from potential harm; >65% of parents say they want an increased sense of control" on IG); Ex. 326 2 at slide 3 ("Caregivers want support and education from Instagram on features and tools that can help 3 them keep their teens safe.") (emphasis in original). 4 (6) Meta's misrepresentations and manipulations 5 Meta's failure to exercise reasonable care is compounded by a series of outright lies—which 6 prevented even the most vigilant administrators, teachers, parents, and students from understanding and 7 heading off the dangers inherent to Instagram and Facebook. We have already discussed Meta's false 8 statements to Congress about its knowledge of addiction and anxiety, as well as its publication of false 9 data underreporting the prevalence of problematic use. See supra § III.A.2.a.4. 10 Meta also lied to Congress about its goals—claiming, falsely, that it has not set out to increase the 11 time users spend on its platforms. Ex. 11A at 171:12-16 (Q to Mr. Zuckerberg: "So three times before 12 Congress you've testified under oath that you don't give your teams goals of increasing the amount of 13 time people spend on your platforms, correct?" A: "Yes.") & Ex. 149, Ex. 150, Ex. 151 (video of 14 testimony). In truth, Meta has not only set quantitative "time spent" goals, it has set those goals for teen 15 users in particular. Ex. 152 at 2613-14 ("Q1 goals" in 2014 include "Time spent. Grow by 5% for H1 16 (across all interfaces, including Instagram"; Zuckerberg: "The most concerning of these to me is time 17 spent."); Ex. 153 at 3741 & Ex. 11A at 178:15-17 (re: 2016 document - Q: "the second goal is time spent 18 increases by 12 percent, correct?" A: "Yes."); Ex. 154 at 4688, 4692 ("2017 Teens Strategic Focus" 19 document sent to Zuckerberg: "Strategic goal: Time Spent Share") & Ex. 11A at 181:17-182:12 (Q: "And 20 there's an example of that underneath, right, 15 billion minutes per day across the Facebook family?" A: 21 "Yes."); Ex. 155 at 2304 (H2 2017 Instagram "Goals" include "US Teen DAP Accounts" of "12.1M" and 22 "Time Spent" of "28 mins") & Ex. 11 at 184:15-186:9; Ex. 156 at 0881 (H1 2019 Facebook document: 23 "Time spent (per DAP) is showing sustained growth and is ahead of H1 goal"); Ex. 157 at 3524 (outlining 24 targets for Instagram of "40 minutes per DAU * 1.5B DAU" in 2023, growing to "46 minutes per DAU * 25 1.95B DAU" in 2026); Ex. 158 at 0347 (Adam Mosseri: "My stretch goal was to get on track to pass 26 TikTok in terms of time spent worldwide"); Ex. 159 at 7078 & Birnbaum Dep. 269:22-270:20 (Meta's 27 expert: Q: "as of the second half of 2023, Facebook was measuring business value through a metric called 28 revenue weighted time spent, yes?" A: "That seems to be the case."). As one Meta employee put it: "No 68 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 81 of 235 1 one wakes up thinking they want to maximize the number of times they open Instagram that day. But 2 that's exactly what our product teams are trying to do." Ex. 226 at 1686; see also Ex. 330 at 1441 3 (LinkedIN founder Reid Hoffman to Mark Zuckerberg: "I've chatted with a few ex-FBers…they all (to a 4 person) told me that FB culture is emphasizing sessions (#, length) over everything else and that they felt 5 that this interpretation of 'emphasizing profit over people' was accurate"). 6 Meta has also lied repeatedly about the existence of dangerous content on its platforms. That 7 includes images and videos glorifying eating disorders, self-harm and suicide, and it even includes child 8 pornography (which, the Court needs no reminding, enjoys no First Amendment protection as it "is 9 intrinsically related to the sexual abuse of children," New York v. Ferber, 458 U.S. 747, 759 (1982)). To 10 be clear, the Districts do not contest that Section 230 provides Meta with the latitude to publish, or de- 11 publish, whatever content it pleases on Instagram and Facebook (although the possession and distribution 12 of child pornography is a federal crime, e.g., 18 U.S.C. § 2252). But nothing in the law gives Meta the 13 right to lie about its choices to children and their protectors. See, e.g., Barnes v. Yahoo!, Inc., 570 F.3d 14 1096, 1107 (9th Cir. 2009), as amended (Sept. 28, 2009) (breaking a promise is not protected by Section 15 230); Doe v. Internet Brands, Inc., 824 F.3d 846, 851 (9th Cir. 2016) (failure to warn is not protected by 16 Section 230). 17 In early 2019, Meta responded to the death of Molly Russell, a fourteen year old British schoolgirl 18 who, according to an official investigation of the London coroner, "died from an act of self-harm whilst 19 suffering from depression and the negative effects of on-line content." Ex. 161 at 2; see id. (coroner: "The 20 platform operated in such a way using algorithms as to result, in some circumstances, of binge periods 21 and images, video clips and text some of which were selected and provided without Molly requesting 22 them."). Meta issued a statement stating, in part, "We do not allow content that promotes or encourages 23 eating disorders, self-harm or suicide and use technology to find and remove it." Ex. 160 at 23; see Ex. 24 180 (Feb. 2019 statement by Mosseri). Meta also updated its public-facing policies to state, "we do not 25 allow people to intentionally or unintentionally celebrate or promote suicide, self-injury, or eating 26 disorders" and "[w]e remove any content that encourages suicide, self-injury or eating disorders…and any 27 self-injury content which is graphic regardless of context." Ex. 162 at 1 & Ex. 163 at 1; Ex. 2A at 218:16- 28 20. Meta executives also appeared before the Senate, testifying, "[w]e don't recommend…eating disorder- 69 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 82 of 235 1 related content to people of any age," Ex. 164 at 27 (Adam Mosseri, Dec. 8, 2021) and "we do not direct 2 people toward content that promotes eating disorders…We actually use AI to find content like that and to 3 remove it." Ex. 165 at 29 (Antigone Davis, Sept. 30, 2021). These are lies. 4 Start with eating disorder content. It is true that Meta relies, in part, on AI tools (which it calls 5 "classifiers") to detect content that violates its eating disorder policy. Ex. 166 at 16-17. But it is false that 6 Meta uses AI to remove that content. In this litigation, Meta has admitted "under penalty of perjury," id. 7 at 35, that its eating disorder classifiers "do not currently perform automatic deletions," id. at 16. Further, 8 "therefore there is no threshold value for autodeletion," id.—meaning, even if Meta's AI tools determine 9 with 100% confidence that a photo or video violates its eating disorder policy, Meta will not automatically 10 delete it. Id. at 16-17; Ex. 985 ¶ 106 (expert report of Prof. Arvind Narayanan). Not surprisingly, and 11 contrary to Meta's representations, Instagram and Facebook are awash in content that promotes eating 12 disorders. According to a 2019 survey of 30,000 Instagram users (which Meta never disclosed), 18% 13 reported viewing content that promotes eating disorders or unhealthy weight loss in the previous week. 14 Ex. 168A at 123:25-124:5; Ex. 101 at slide 6. As of 2020, it remained the case that "it's easy to find 15 violating ED accounts in a couple of taps. When I search for 'thinspo' or 'proana' the accounts shown are 16 overwhelmingly violating." Ex. 179 at 6755. Meta knew this was "us not delivering on the promise made 17 by Adam in February 2019." Id. at 6751. 18 The situation is just as bleak with respect to content promoting self-injury and suicide. Again, Meta 19 uses AI classifiers to detect this content. But again, it does not use those classifiers to take reasonable and 20 prompt action. Meta's under-oath discovery responses reveal that, even if its classifiers are 93.99% certain 21 that photos and videos glorify self-harm, Meta will not automatically remove that content from its 22 platforms. Ex. 166 at 13 ("Threshold value for autodeletion: 0.94."). Internally, Meta employees 23 understand that "violating content" can "score[] significantly lower than th[is] established threshold." Ex. 24 182 at 8895 (admitting this is a "root cause" of the "problem"). Again, the results are predictable: "There's 25 going to be SSI [suicide and self-injury] content on IG and FB for the near future. The existing classifiers 26 are not hard to game." Ex. 171 at 6015 (internal chat from 2019); Ex. 172 at 0021 ("Suicide Video 27 Producers and Suicide Promoters have some very high reach accounts"). As of 2018, Meta knew that, 28 "[o]n a daily basis, 5.1 million users, or 0.75% of DAP, are exposed to SSI-related content." Ex. 178 at 70 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 83 of 235 1 4409. As of 2021, it knew (through a survey of nearly 250,000 users) that 6.7% of respondents were 2 exposed to self-harm or suicide on Meta's platform in a given week. Ex. 145 at 5049 & Ex. 6 at 506:6- 3 507:11; see also Ex. 146B at 441:7-11 (survey was "massive even for Meta survey standards"); Ex. 3 at 4 241:15-18. 5 This survey also found that the users most exposed to this content are the youngest—with 8.4% of 6 13- to 15-year olds reporting having seen "someone harm themselves, or threaten to do so, on Instagram" 7 in the past week. Ex. 145 at 5033, 5049. It was "well known among the well-being team at Meta … that 8 suicide and self-injury and self-harm content was easily accessed on Instagram by kids." Ex. 168 at 9 346:13-347:1. "We know that SSI and ED have a significantly disproportionate large teen audience." Ex. 10 177 at slide 11 (Oct. 2023). "Everyone involved with SSI and eating disorders," including "producers" 11 and "reporters," "are more likely to be in early high school or late high school." Ex. 172 at 0030. 12 Meta executives knew there was a problem. Ex. 174 at 5152 (Monika Bickert, 2019: "IG's product 13 is essentially violating our content policies by displaying self-injury images without captions."); Ex. 175 14 at 5341 (Nick Clegg, 2019: "I said publicly…that we would do 'whatever it takes' to address the legitimate 15 concerns raised….This lack of connection between what we commit to publicly and what we do internally 16 really worried, and slightly shocked, me."); Ex. 182 at 8889 (April 2019: over 315,000 daily active people 17 exposed to SSI content through algorithmic recommendations on Instagram Explore); Ex. 179 at 6750 18 (2020: "[with] the inquest into Molly Russell's death…the focus will be on what we were recommending 19 Molly – and most crucially whether we have done enough to prevent 'future deaths.'….On search we're 20 exposed with nowhere to hide."); Ex. 168A at 269:20-270:8. They were informed by an audit committee 21 that "Instagram does not appear to adequately remove, label, or demote pro-restrictive eating disorder 22 content, nor pro-suicide and/or pro-self-harm materials." Ex. 183 at 0074. They were even informed by 23 parents (via a Meta-sponsored focus group) that they should "alert parents, schools, [and] emergency 24 services" about this issue. Ex. 181 at 4759; see id. at 4758 ("It absolutely confounded them that we could 25 be working on an issue so long, yet we are where we are."). 26 Characteristically, that is not what Meta chose to do. Instead, it maintained a "reactive" posture to 27 removing SSI content. Ex. 182 at 8907 (3/4 of SSI content flagged reactively rather than proactively); Ex. 28 176 at 4022 (with respect to "viral suicide & self-injury challenges," "[w]e've historically taken a reactive 71 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 84 of 235 1 approach … (i.e. we deal with it once it's become so viral that media outlets catch wind of it."); Ex. 2A 2 at 255:5-256:21. It rolled out new features on which SSI content couldn't be flagged, let alone removed. 3 Id. at 48:15-49:16, 53:2-20 (explaining how classifiers were "very ineffective on short-form video," a 4 problem that Meta "[w]as not able to fix before the launch of Reels."). And it published misleading 5 statistics minimizing these problems. Compare Ex. 147 at 1 (public-facing "Community Standards 6 Enforcement Report" estimating that "the upper limit" of content violating Meta's suicide and self-harm 7 policy was "0.05%.") with Ex. 145 at 5049 (internal survey estimating 6.7% exposure to SSI). As one 8 former employee acknowledged, Meta's public data presented a "deliberately incomplete picture in an 9 underreporting of the rate of harmful content and experiences that were happening on our platform." Ex. 10 3 at 457:22-458:1; see also Ex. 2A at 222:6-12 (acknowledging disconnect between Meta's SSI policy 11 and the aggregated content recommended by Instagram). 12 Most shockingly of all, Meta has lied about its approach to removing child pornography on its 13 platforms. Since 2015, Instagram's Community Standards have stated, "We have zero tolerance when it 14 comes to sharing sexual content involving minors or threatening to post intimate images of others." Ex. 15 184 at 11. This is unquestionably the right thing for Meta to say. But it is also unquestionably false. Meta 16 operates three AI tools (or "classifiers") meant to detect content that likely violates its policies against 17 child sexual abuse material ("CSAM"), CSAM solicitation, and child sexualization, respectively. Ex. 166 18 at 29-30. One would hope that, if these tools determine content more likely than not violates these policies, 19 it would be automatically removed. Meta does not take such an approach—not even close. To the contrary, 20 even if these classifiers determine with 100% confidence that photos or videos violate its policies against 21 CSAM and child sexualization, Meta will not automatically delete that content. Id.; Ex. 985 ¶ 106. This 22 was an intentional choice. In 2021, Meta product managers acknowledged they "could reduce [the] 23 precision thresholds for quarantine or auto-delete" of content violating its child safety policies. Ex. 188 at 24 5478. (Translation: Meta could go from auto-deleting none of the content flagged by these classifiers to 25 at least auto-deleting some of it.) But they decided against doing this, citing a "high risk" of false positives, 26 "harsh penalties for child safety offenses," and a "[l]ack of appeals staffing (so inability to correct account- 27 level disables"). Ex. 188 at 5478. It goes without saying that Meta could have solved this problem by 28 simply hiring more appeals staff. Instead, it took the approach most likely to keep known child 72 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 85 of 235 1 pornography on its systems. This is the opposite of zero tolerance. See also Ex. 189 at 1907 & Ex. 2 at 2 371:13-373:20 (interstitial allowing users to "see results anyway" in response to search query that "may 3 be associated with child sexual abuse"). 4 Meta's human content moderation for child sexual abuse material has been just as deficient as its 5 automated systems, and Meta has also lied about its efficacy. In March 2020, Mr. Zuckerberg told 6 reporters that Meta was "actually surging the number of people who are working on" moderating "the 7 most sensitive types of content," including "work on child exploitation." Ex. 187 at 3403. This was another 8 lie. Ex. 2 at 357:11-23 (Q: "in fact, Meta was not surging the number of people working on child 9 exploitation, correct?" A: "Not the total number of people, no." Q: "And as a consequence of that, the 10 capacity to review potential child sexual abuse material on Instagram went down significantly during 11 COVID; isn't that correct?" A: "Yes, that's correct."). Indeed, as of November 2020, Meta's "human 12 review capacity" for child exploitation imagery was "at ~30%" of what it needed to be—with "<20% of 13 projected supply for the near future." Ex. 187 at 3402 & 357:18-358:14; see also Ex. 186 at 9785 (2021: 14 "Up until last half…no workstream set up to do bulk enforcement against FB Groups with _clearly_ 15 violating content in either CEI [child exploitation imagery] or MS [minor sexualization]") 16 The consequences of Meta's upsettingly high tolerance for child pornography and sexualization 17 have been devastating. See Ex. 2A at 272:3-9; Ex. 185 at 1113 (2020: "Is anyone else receiving child 18 abuse reports on the platform? In the last 24 hours I have received a ton… I've seen terrible things, we're 19 not detecting it."); Ex. 210 at 2309 ("21x fewer" deletions of child exploitation imagery on Instagram 20 compared to Facebook), 2310 (over 10,000 users following known child exploitation related hashtags 21 such as "#sexykid" and "#13yearoldboy"), 2311 (over 27,000 threads between bad actors and a teen over 22 a 7 day period), 2314 (providing examples of "accounts sexualizing teens"). See also id. at 2324 & Ex. 2 23 at 302:3-303:20 (simply typing "h" into search bar could return results including "hottieskids" and "hot 24 young boys"), id. at 2327 & Ex. 2A at 305:16-306:16 (searching for "childporn" returns numerous hashtag 25 results with those words, including "#childporncollection" and "#childpornisgood"); Ex. 325 at 0324 & 26 Ex. 2A at 286:16-25 (for test account created by Meta, "CEI being recommended as the top result in the 27 Explore section"). 28 73 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 86 of 235 1 The examples discussed above reflect the reality that child safety is simply not a priority at Meta— 2 as additional internal documents make explicit. See Ex. 363 at 0155 ("Child Safety is explicitly called out 3 as a non-goal in our H2 plans."); Ex. 378 at 9908 (saying that Meta's efforts on child grooming are 4 "somewhere between zero and negligible" and adding "Child safety is an explicit non-goal this half"); Ex. 5 1068 at 5137-38 (showing there were five engineers, two data scientists, and one data engineer to cover 6 "all child safety problems" including "the multiple other problems that [Meta] have to 'non goal' on – no 7 sexual child abuse – child nudity"). 8 At no point did Meta warn parents that it had only a fraction of the bodies it needed to review 9 potential child sexual abuse material on Instagram. Ex. 2B at 358:6-21. Indeed, as recently as January 10 2024, Senator Josh Hawley asked Mr. Zuckerberg if Meta's platforms were "a disaster for teenagers" and 11 Mr. Zuckerberg responded: "Senator, that's not true. That's not true." Ex. 467 at 37. But it is true. The 12 same day Mr. Zuckerberg testified, Meta's employees were declaring a "SEV" (sitewide emergency event) 13 related to child safety—after a team "discovered that a number of Child Safety banked search terms" were 14 not being applied to block results on Instagram Search. Ex. 468 at 1412. 15 Meta has never publicly corrected its zero tolerance policy or otherwise warned school counselors, 16 teachers, parents, or kids that its policy was not being enforced, manually or automatically. Ex. 2A at 17 313:21-25 (acknowledging inconsistency), Ex. 2A at 314:2-315:17, 474:18-22. Nor did Meta warn the 18 public that its algorithm sometimes recommends child pornography to Instagram users. Ex. 2 at 262:10- 19 263:4 (Q: "In your experience, did Instagram's algorithm, in fact, sometimes recommend to its user sexual 20 content involving minors?" A: "Yes."), 286:16-25 (Instagram's internal investigation indicated that 21 CSAM was being recommended in Explore). Instead, Meta continues to falsely maintain that "[o]ur 22 systems are effective at reducing harmful content." Ex. 204 at 4 (quote in Wall Street Journal); compare 23 id. with Ex. 2B at 381:23-382:3 (Q: "Based on your experience working at Meta for three and a half years 24 on child safety issues, do you think it's accurate to say that Meta's systems are effective at reducing 25 harmful content related to child sexualization and child exploitation?" A: "No."). Meta's willful and 26 knowing disregard of the rights of others. 27 (7) Meta's willful and knowing disregard of the rights of others 28 Meta has a unique responsibility to protect the youngest users of its platforms, as its own 74 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 87 of 235 1 employees have acknowledged. Ex. 379 at 7816 ("we have more responsibility to protect teens"). Instead 2 of embracing that responsibility, Meta has prioritized its own profits—over and over and over again. The 3 preceding pages show how this issue played out, over years and across initiatives. 4 The Districts have submitted an expert report from a corporate ethicist explaining how Meta 5 violated its own code of conduct and engaged in "pseudo corporate social responsibility." See Ex. 993 ¶¶ 6 156-388. But testimony from Meta's own employees says it all. Vaishnavi Jayakumar, former Head of 7 Safety and Well-Being at Instagram from 2020-2023, testified that it was "generally pretty difficult to 8 make safety changes that might impact growth or daily active users by any significant amount." Ex. 2B at 9 384:22-385:2. Dr. Joshua Simons MP, a research scientist at Facebook AI from 2018-2022, testified that, 10 "if we were proposing something that would reduce engagement that went up for the executive team, 11 Mark Zuckerberg, to review, the decision that came back would prioritize the existing system of 12 engagement over other safety considerations." Ex. 13 at 60:15-61:11. a senior UX 13 researcher at Meta from 2017-2022, regretted that she "joined the company a little naïve about what the 14 company would want to do" and that research-backed recommendations "were rejected because the 15 changes would have limited certain metrics, like engagement." Ex. 10 at 34:19-23. Brian Boland, whose 16 eleven-year tenure at Meta ended as VP of Partnerships, testified that, "My feeling then and my feeling 17 now is that they don't meaningfully care about user safety. It's not something that they spend a lot of time 18 on. It's not something that they think about. And I really think they don't care." Ex. 8 at 109:2-16. 19 These are individuals from different parts of the organization, who worked there at different times, 20 with varying degrees of seniority. Their perspectives are entirely consistent. And they are not alone. See 21 also, e.g., Ex. 3 at 460:21-461:5 (Allison Lee: "they perhaps cared as a third or fourth order priority about 22 safety and well-being"); Ex. 23 at 247:7-10 (Frances Haugen: "ongoing trend of not prioritizing safety as 23 high as it should have been"); Ex. 168B at 482:20-21 (Arturo Béjar: "prioritize[d] engagement and growth 24 over safety and well-being."); Ex. 202 at 1 (Sarah Wynn-Williams: leadership "maintained their focus on 25 growth" even though "[t]hey knew the negative externalities of their products were being pushed onto 26 teens"); Ex. 256 at 59:8-20 (Lotte Rubaek: child safety "not at all" a priority). 27 Despite earning billions of dollars in annual revenue—and its leader being one of the richest 28 people in the world—Meta simply refused to invest resources in keeping kids safe. One final story sums 75 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 88 of 235 1 this up. In April 2017, Mr. Systrom wrote to Mr. Zuckerberg appealing for additional staff to address a 2 variety of safety issues as to which "women and teens [are] particularly impacted." Ex. 663 at 6745. Mr. 3 Systrom emphasized the importance of upholding his "public commitment to making Instagram a place 4 where people feel safe." Ex. 663 at 6744. Mr. Zuckerberg responded: "we're facing more extreme issues 5 on FB right now….I do view funding that as more urgent in the near term." Id. He continued, "I probably 6 can't get you 13 engineers in the near term, even though I'm supportive of these projects." Id. Mr. Systrom 7 groused to his team: "I don't think Mark understands the urgency of working on integrity related issues at 8 IG." Id. at 6751. 9 By October 2017, the situation had not improved. According to an internal quarterly review, 10 Instagram's protect and care (PAC) team remained "far behind" that of Facebook—"we could become a 11 major liability for FB Inc." Ex. 666 at 1333. "Instagram is now over 850M monthly users, and with that 12 size comes the responsibility to ensure our community is safe. Also, we're launching more new products 13 at a faster rate than ever, each of which require additional work to prevent abuse and reduce exposure of 14 FB Inc. We're continuing to see an increase in high intensity abuse." Id. at 1334. "However, the PAC 15 team is only 22 engineers. We aren't staffed …to stay ahead of all potential PAC related issues." Id. 16 Fast forward two years. On April 8, 2019, a senior executive at Meta wrote to Mr. Zuckerberg 17 informing him that the company had a "deep understanding around three negative drivers that occur 18 frequently on FB and impact people's well-being," including "problematic use (prevalence: 55% mild, 19 3.1% severe)" and "social comparison (prevalence: 40% mild, 5% severe)." Ex. 398 at 5473. The 20 executive asked Mr. Zuckerberg for an additional headcount of 24 heads, to fund a "Well-being 10x" 21 initiative. Id. "We believe there is a strong need to increase our investment in these areas to make a 22 meaningful shift in these areas." Id. Mr. Zuckerberg did not respond. But Adam Mosseri did: "I don't see 23 us funding this from Instagram any time soon." Id. "The 24 head count were not funded" and the team 24 was asked "to do the work with existing head count that they have." Ex. 399 at 226:15-18. 25 Fast forward another two years. On August 27, 2021, Meta's Head of Global Affairs, Nick Clegg, 26 wrote to Mark Zuckerberg with a "proposal…for additional investment to strengthen our position on 27 wellbeing across the company." Ex. 278 at 0952. "Our wellbeing work is both under-staffed and 28 fragmented across teams," he wrote. "We are not on track to succeed for our core well-being topics," 76 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 89 of 235 1 including "problematic use … and SSI." Id. "[I]f not addressed, these [problems] will follow us into the 2 Metaverse." Ex. 278 at 0952. Again, Mr. Zuckerberg did not reply. 3 Two months later, in October 2021, Arturo Béjar wrote to Mr. Zuckerberg. Mr. Béjar had served 4 as the head of Facebook's Protect and Care team for years and was then working as a consultant for Meta. 5 He warned Mr. Zuckerberg that, per internal research, "51% of Instagram users say 'yes' to having a bad 6 or harmful experience in the last 7 days" and nearly "39.4% of 13-15 year olds said they experienced 7 negative [social] comparison." Ex. 303 at 4804-05. Mr. Béjar went on to identify well-being solutions and 8 tell Mr. Zuckerberg a variant of what other executives had been saying for years: "I believe that it is 9 important to get the[se] efforts well-funded and prioritized." Id. at 4805. Mr. Zuckerberg did not reply. 10 That same week, Frances Haugen publicly testified before Congress about hundreds of internal 11 research documents that Meta had kept secret for more than a decade, many concerning the impact of 12 Instagram and Facebook on youth mental health. Meta responded by sending Mr. Mosseri to the Hill to 13 placate Congress. Mr. Mosseri testified: "As Head of Instagram, I am especially focused on the safety of 14 the youngest people who use our services." Ex. 164 at 2. But behind closed doors, Mr. Zuckerberg was 15 texting confidants, "I'm not going to say [child safety] is my personal main focus when I have a number 16 of other areas I'm more focused on like building the metaverse." Ex. 664 at 2509. 17 In November 2021, Nick Clegg renewed his efforts to acquire additional funding for well-being 18 work, sending a "revised investment proposal" that "scal[ed] down" his initial funding request. Ex. 278 at 19 0951. "Given the increased urgency of all this since the initial escalation in August, are you supportive of 20 funding this scaled back proposal?" he asked. Id. Despite this plea, no additional headcount was 21 forthcoming. Ex. 278 at 0950; Ex. 9B at 472:16-17. Mr. Zuckerberg's response? Silence. 22 b) TikTok's Conduct 23 (1) TikTok's business model 24 TikTok, like its co-defendants, built its extraordinary success on one commodity: attention. Its 25 business model depends on capturing and holding users—especially school-aged children—for as long as 26 possible, because every additional minute on the platform translates directly into revenue. Together with 27 its parent company, ByteDance, TikTok monetizes that attention through what its own employees 28 described as an "advertising business model" that "create[s] incentives for companies to grow at the 77 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 90 of 235 1 expense of its users." Ex. 632A at 50:5-10, 103:16-17. Executives have acknowledged that advertising 2 provides the vast majority of TikTok's revenue. Ex. 636 at 47:11-21. Internal documents confirm the 3 priority: , and 4 , set out to "[a]ccelerate ads revenue to " in a single quarter of 2023— 5 up from 2022's per day, or roughly annually. at 47:3-10; Ex. 522 at 6 6223; at 451:1-6; Ex. 523. 7 declared the corporate governing principle that would 8 drive TikTok's every decision: "With more active users, more clicks, and better user satisfaction comes 9 higher traffic, which is then turned into revenue." Ex. 434 at 7156; at 115:7-117:15. That was 10 not an aside—it was a directive. Inside TikTok, the connection between user engagement and profit was 11 not merely understood; it was institutionalized. An internal document—confirmed by TikTok's former 12 —put it bluntly: "the advertising-based 13 business model encourages optimization for time spent in the app." Ex. 541 at 9176; at 60:15- 14 61:1. The logic was as simple as it was deliberate: more time meant more ads, and more ads meant more 15 money—billions! 16 TikTok pursued that attention-for-profit model relentlessly. Internal documents show that in just 17 one quarter of 2023, the company targeted daily active U.S. users. Ex. 477 at 44:20- 18 45:16; Ex. 522 at 6223. "Stay duration"—the time users spent in the app—was treated as a "top-level 19 metric." Ex. 635 at 410:15-19; Ex. 525. The mandate was clear: increase average use from per 20 day in 2021 to in 2022. Ex. 635 at 455:9-23; Ex. 526. To achieve that goal, TikTok deployed 21 what its own records describe as "powerful coercive design tactics" that "tend to benefit companies and 22 advertisers more than users." Ex. 632A at 103:20–104:10; Ex. 541 at 9185. These were not accidents of 23 user behavior—they were deliberate design choices engineered to maximize time, attention, and revenue. 24 Even TikTok's supposed "safety" features were judged by a single yardstick: whether they hurt 25 engagement—not whether they protected kids. When TikTok tested weekly screen-time updates for 26 minors, the internal takeaway was that the feature had "neutral" effects—no change in daily active users, 27 stay duration, or retention. Ex. 405A at 199:13–200:5; Ex. 488 at 7711. Another internal report, tracking 28 a "Screentime Management Upsell for Minors," listed "Stay Duration" as a "Guardrail Metric" and 78 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 91 of 235 1 proudly noted "no statistically significant change." Ex. 405A at 203:6-204:11; Ex. 487 at 7234. The irony 2 could not be clearer: these so-called guardrails were never meant to protect children—they were meant to 3 provide the appearance of safety while ensuring that neither engagement nor revenue were ever 4 meaningfully reduced. In short, "safety" was acceptable only if it left the bottom line untouched. When 5 one test showed a decline in stay duration, the report reassured executives that the drop "align[ed] 6 within guidance" and had "no impact to retention." Ex. 635 at 431:3-15; Ex. 497 at 2841. That same report 7 was blunter still, acknowledging "minors do not have executive mental function to control their screen 8 time." Ex. 497 at 2841. 9 The picture could not be clearer. TikTok's business model does not tolerate limits on user 10 attention—it thrives on eradicating them, even at the expense of safety. Every design choice, platform 11 tweak, and so-called "safety" feature was measured by one metric: did it keep users, especially children, 12 on the platform longer? The longer the stay, the more ads TikTok served, and the more revenue flowed to 13 ByteDance. That was no accident. It was the overarching strategy—engineered, enforced, and embedded 14 in the corporate culture. What mattered most was not safety, not user well-being, but screen time. 15 (2) TikTok's targeting of school-aged children 16 From the start, TikTok identified school-aged children—particularly middle and high school 17 students—as its core audience. In 2018, after ByteDance acquired Musical.ly and rebranded it as TikTok, 18 internal documents show that ByteDance and TikTok knew a large share of their users were "middle 19 school and high school females." Ex. 411 at 63:2-16, 65:12-66:5, 66:10-66:21; Ex. 412 at 5197. Rather 20 than distancing themselves from this young audience, they embraced them—formally defining school- 21 aged children as the platform's "core audience." at 116:18–25; Ex. 414 at 0624 (former 22 admitting that "high schooler[s] is a core audience of our platform"). TikTok made clear it 23 wanted school-aged children to be its heaviest users, announcing a "rebrand" aimed at ensuring those ages 24 13 to 18 would be "heavy user[s]." Ex. 415 at 0205. Internal planning documents went further, identifying 25 "high school boys" as a "[k]ey population group[]." Ex. 422 at 0220A; Ex. 413B at 109:14–20. 26 To better exploit this demographic, TikTok launched an internal initiative—"Project Cool Kid"— 27 to study and capture the habits of school-aged users. The project focused on "how High Schoolers, 28 predominantly L2 [15-17-year-old] males are using TikTok," to "understand high school trends," and to 79 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 92 of 235 1 learn what "fuels usage and adoption in High Schools." Ex. 423 at 7326; Ex. 411 at 132:6-134:11. TikTok 2 and ByteDance wanted to know "how mainstream TikTok [was] in high school and whether or not all 3 high schoolers are using TT." Ex. 424 at 9792; Ex. 413B at 127:18–24. 4 As part of this research, TikTok conducted "street interviews" with middle and high school 5 students. Ex. 411 at 166:17-168:12. One , reported: "dude I talked 6 to around 30 high schoolers," and boasted, "literally all middle school to high schoolers, like 8–18, have 7 heard of TikTok in Houston and Chicago." at 164:1-12, 176:10-13; Ex. 425 at 2374; Ex. 426 at 8 5153. These interviews were not confined to junior staff. Top executives from both ByteDance and 9 TikTok—including , 10 and —personally walked the streets conducting interviews with school-aged children. 11 at 165:10-166:14. TikTok conducted these interviews annually. at 184:19-21. The findings 12 were telling: children as young as eight used TikTok "every day," while others admitted that "everyone 13 says you'll get addicted." Ex. 411 at 186:10-15, 211:5-22; Ex. 428 at 5042; Ex. 429 at 14. 14 Armed with this data, TikTok set out to capture the group of adolescents it had a thin hold on from 15 Musical.ly—high school males. It launched targeted advertising campaigns on co-defendant Snap's 16 platform, Snapchat "with the aim to grow L2 [high school-aged, 15-17-year-old] male users," shifting ad 17 spend from Facebook to reach a younger audience. Ex. 412 at 5220; Ex. 411 at 109:22-110:23. TikTok 18 quickly tripled its daily ad spending to approximately 19 Ex. 508 at 269:24-270:11; Ex. 430 at 3467. They even removed "age filters on paid [ad] channels 20 in the U.S. market" to maximize reach among teenage males. Ex. 508 at 137:2-139:3; Ex. 431 at 8836. 21 These efforts were central to TikTok's 2019 objectives and key results: to "attract teen male users" and 22 "drive market growth." Ex. 413B at 111:15-23; Ex. 422 at 0220A. 23 The strategy worked. By the second half of 2019, the percentage of 15–17-year-old males among 24 daily new users "grew significantly...reaching and surpassing all other user groups except for L2 25 [high school] female[s]." Ex. 411 at 113:21–114:3; Ex. 412 at 5221. Buoyed by this success, TikTok 26 declared itself "mainstream" in high schools. Ex. 411 at 124:19-125:6; Ex. 412 at 5196, 5220-5221. Even 27 as it later sought to "age up" its user base, TikTok acknowledged it was "maxed out" with middle and 28 high school users—and committed internally to not "eliminating or overlooking" users ages 13–17, the 80 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 93 of 235 1 core of its user base. Ex. 411 at 200:7-21, 373:25–374:5; Ex. 432 at 7167; Ex. 433 at 156:3-7. 2 In September 2021, TikTok celebrated its own success with a video press release titled "Thanks a 3 Billion," boasting that "more than 1 billion people from around the world now come to TikTok every 4 month." The announcement featured , the same executive who had walked the 5 streets of American cities interviewing students. See Ex. at 135:13-20, 136:3-137:20; Ex. 569. 6 TikTok has been staggeringly successful at attracting young people to its platform. In sworn 7 discovery responses, TikTok confirmed that users ages 14–18 skyrocketed from in 2018 to 8 more than by 2022 in 2024: 9 Year Users ages 14-186 10 11 12 13 14 Ex. 410 at 22–23. These figures confirm that TikTok's campaign successfully entrenched its dominance 15 among middle and high school students—making the resulting harms to schools not only foreseeable but 16 inevitable. By June 2022, TikTok had achieved a penetration rate among children aged 13–17. 17 Ex. 433 at 141:1-6; Ex. 538 at 6. As , admitted, a penetration rate 18 was . at 151:13-152:3. 19 With that dominance came advertisers—and money. ByteDance's 20 , made TikTok's priorities clear: to acquire and retain users while "[g]enerating revenue." 21 at 118:18-119:7; Ex. 434 at 7155-56. TikTok's year-end business reviews echoed that mandate. 22 The 2021 review set a target of billion in advertising revenue and noted that average daily time 23 spent on the app had climbed to minutes. Ex. 437 at 134:5-9; Ex. 435 at 7029. Internal records confirm 24 the underlying philosophy: an "advertising-based business model [that] encourages optimization for time 25 spent in the app." Ex. 436 at 9176. 26 27 6 TikTok did not provide statistics that included 13-year-olds, omitting a significant portion of its underage 28 user base. 81 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 94 of 235 1 To sustain its growth, TikTok marketed its young audience directly to advertisers. at 2 203:3–24 ( , confirming that "Gen Z was a very—a very 3 common topic of conversation with advertisers"); Ex. 438 at 3, 5 (noting of users are Gen Z and 4 that "Gen Z watch and interact with more videos per day"). In 2019, TikTok's brand partnership deck 5 touted explosive U.S. user growth and described the platform as "Generation Z Focused." Ex. 437 at 6 99:25-100:2; Ex. 439 at slide 14. The same presentation emphasized that users opened TikTok more than 7 daily and spent over a day on the app. Ex. 437 at 103:1-7; Ex. 439 at slide 15. 8 TikTok made similar appeals to individual brands. In a presentation to , TikTok 9 reported that of its U.S. audience was between ages 13 and 24, highlighting "sustained engagement" 10 and admitting that "our algorithm makes [the] app very addicting." Ex. 437 at 212:22-213:5; Ex. 440 at 11 9376, 78-79. In a pitch to , TikTok boasted that of its audience was between ages 13 and 12 17— —and that this cohort spent the most time on the platform, averaging 13 more than per day. Ex. 437 at 224:8-10; Ex. 441 at 9049–9050. TikTok also emphasized that 14 Gen Z users were " ," reporting that of its 15 monthly active users were ages 13-17 as of April 2021. Ex. 437 at 230:10-231:5; Ex. 442 at 19. 16 The record leaves no doubt: TikTok deliberately targeted school-aged children, studied their 17 habits, and built its platform around keeping them engaged. Its own data, internal communications, and 18 marketing materials confirm this was not incidental—it was the primary business strategy. 19 (3) TikTok's targeting of schools 20 To cement its dominance among school-aged children, TikTok turned its attention to schools 21 themselves—viewing them as the most direct path to widespread adoption. It sought to embed TikTok in 22 the educational ecosystem, partnering with national organizations and developing initiatives that reached 23 directly into classrooms. Like Meta, Snap, and Google/YouTube, TikTok borrowed from Big Tobacco's 24 playbook—using "youth prevention" and school programs to infiltrate classrooms and burnish its image 25 under the guise of education. Every strategy served a single goal: making TikTok an unavoidable part of 26 students' daily lives. 27 TikTok began by disguising commercial objectives behind partnerships that appeared educational 28 or community-driven. Chief among these was its collaboration with the National Parent Teacher 82 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 95 of 235 1 Association ("National PTA"), a respected organization representing millions of students, parents, and 2 educators. In 2019, TikTok initiated what it described as a "formal relationship" with the National PTA, 3 leveraging its reputation to gain access and legitimacy within schools. Ex. 413B at 182:7-183:11. 4 Internally, TikTok described the PTA partnership as an opportunity to "leverage" its vast 5 network—54 state PTAs, 8,000 council PTAs, and 24,000 local PTAs, representing 4 million members 6 and 16.2 million families. Ex. 443 at 4129, 4133. TikTok identified several advantages: (1) the PTA could 7 "positively raise ByteDance's profile among parents"; (2) co-branded materials could reach "PTA leaders 8 and influential, engaged families"; (3) TikTok could "tap[] into key National PTA campaigns and events"; 9 and (4) the company could "draw[] on the community trust, 120-year-old established reputation, and 10 perceived values of PTA." Id. at 4130. 11 Ultimately, TikTok saw the PTA partnership as a vehicle for influence. After securing a paid 12 sponsorship, company officials noted the PTA would "do whatever we want going forward in the fall . . . 13 [t]hey'll announce things publicly[,] [t]heir CEO will do press statements for us. They'll co-brand 14 ConnectSafely's 'How to TikTok Guide.'" Ex. 444A at 50:23-52:4; Ex. 445 at 4722. In October 2019, 15 TikTok executed a Sponsorship Agreement with the National PTA 16 ." Ex. 444A 17 at 58:16-24; Ex. 446 at 8241. This clause , including a 18 in 2020 and an additional in 2024. Ex. 444A at 60:15-61:4, 90:19- 19 94:23; Ex. 447 at 0392; Ex. 448 at 4480. 20 TikTok then used the National PTA as a front to promote its "safety" initiatives, such as "Safer 21 Internet Day." Ex. 444A at 57:4-58:2; Ex. 446 at 8239. It distributed "mass mailer" materials to parents, 22 educators, and school districts promoting TikTok's purported safety features, including the TikTok Guide 23 for Parents and TikTok Family Pairing Guide for Parents. Ex. 444B at 479:24-480:5; Ex. 449; Ex. 450. 24 Internally, it celebrated that "tons of schools" printed the guide and left copies "in the principal's office." 25 Ex. 451 at 4792. 26 These partnerships were not public service—they were public relations. By laundering TikTok's 27 image through a trusted national institution, TikTok sought to reassure parents and educators, deflect 28 scrutiny, and preserve the engagement driving youth harm. 83 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 96 of 235 1 To deepen its reach, TikTok co-hosted "flagship" school events with the National PTA nationwide. 2 Ex. 413B at 162:3-13, 182:9-19; Ex. 448 at 4493. One such event at Westwood High School in November 3 2022 drew 300 parents, students, and administrators. Ex. 452 at 3744. Despite the family-friendly veneer, 4 these events revealed TikTok's knowledge that children under 13 were active users. Ex. 444A at 222:16- 5 228:5; Ex. 453 at 4466 ("This student panel is primarily under 13"). 6 At another event in February 2020 at a Los Angeles school, a student panel featured children under 7 13 who admitted they "lie[d] about their birthdate to get around" TikTok's age restrictions. Ex. 444A at 8 225:7-13, 226:10-20; Ex. 453 at 4466. Internally, executives noted they "seriously dodged a bullet" when 9 CBS and ABC canceled planned national coverage. Ex. 444A at 225:23-226:3; Ex. 453 at 4466. 10 A central goal of the partnership was favorable publicity. TikTok's sponsorship agreements 11 required that TikTok appear on the National PTA's "sponsors and partners" webpage with a link to 12 ByteDance's site and mandated that the PTA provide "positive quotes" about TikTok's sponsorship. Ex. 13 446 at 8241; Ex. 448 at 4495. 14 TikTok capitalized on these commitments immediately. In 2020, TikTok launched a "Trust & 15 Safety" marketing campaign across twelve politically strategic states—chosen for overlap with key 16 congressional committees—and used the PTA to bolster its credibility. Ex. 454 at 7940. The PTA also 17 helped promote TikTok's "Family Pairing" feature, publishing op-eds offering "commentary in support 18 of TikTok's tools and resources," which TikTok called "of enormous value." Ex. 455 at 6205; Ex. 444A 19 at 63:5-64:9. TikTok also used the PTA partnership for political influence. TikTok targeted event sites in 20 "major market media centers" and "sensitive political districts," noting that "Tampa gets us both Senators 21 Rubio and Scott." Ex. 456 at 9201, 9202. Internally, it described National PTA grants funded by TikTok 22 as "[m]ore good news to share with policymakers!" Ex. 457 at 8844. 23 Across all channels, TikTok controlled the PTA's messaging. TikTok employees directly edited 24 PTA quotes and supplied talking points for use with local leaders. Ex. 458 at 6228; Ex. 459 at 2229. The 25 PTA even agreed not to participate in interviews "without first having a planning session" with TikTok's 26 media team. Ex. 460 at 1997. To enforce this control, 27 . Ex. 444A at 158:14-159:5; Ex. 446 at 8237. 28 Beyond the PTA, TikTok paid other educational and "digital safety" organizations to promote its 84 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 97 of 235 1 image. It paid Common Sense Networks "to deliver an education and awareness program" that 2 presented TikTok as empowering teens to "take ownership of its digital lives." Ex. 444A at 115:18-117:5; 3 Ex. 461 at 3873. TikTok cited this partnership in its TikTok Guide for Parents, distributed nationwide 4 through the National PTA. Ex. 444A at 284:16-25; Ex. 449 at 6392. 5 TikTok also paid ConnectSafely, a nonprofit, to "draft and publish a 'Parent's Guide to TikTok'" 6 and co-host events like Safer Internet Day. Ex. 444A at 119:5-16; Ex. 462 at 4071. Internally, TikTok 7 described ConnectSafely's CEO, Larry Magid, as "a good shield for us going forward," able to "serve as 8 a public advocate for [TikTok] in the face of all the criticism." Ex. 444A at 246:7–247:2; Ex. 463 at 1110. 9 TikTok publicly cited the partnership—including in CEO Shou Chew's Senate testimony— 10 . Ex. 444B at 349:04-350:14; Ex. 464 at 3837; Ex. 11 465 at 2626. 12 Taken together, TikTok's network of sponsored partnerships—with the National PTA, Common 13 Sense Networks, and ConnectSafely—was not a safety initiative but a coordinated public-relations 14 campaign. Like Big Tobacco and JUUL, TikTok repackaged corporate promotion as public education, 15 exploiting schools' and parents' trust to entrench its platform among youth. See Ex. 662 (2019 16 congressional hearing regarding JUUL's role in youth nicotine epidemic). 17 TikTok also developed direct plans to enter classrooms, aiming to make TikTok "the #1 short 18 video app for creation and consumption of educational content" and a "resource for classrooms" where 19 teachers and students could "build curricular content on TikTok at scale." Ex. 469 at 0563; Ex. 470 at 20 9753; Ex. 444A at 126:3-25. It envisioned "school-specific Discover Pages" where educators could post 21 content and assign "homework" as TikTok videos, to be piloted in major school districts. Ex. 471 at 6534. 22 To promote this initiative, 23 . Ex. 471 at 6529; Ex. 472 at 9074. These influencers—branded 24 as "Educreators"—were presented as "standout teachers from the TikTok community" 25 to encourage others to post under #TeachersOnTikTok. Ex. 474 at 3472. TikTok sought media 26 coverage of these campaigns, coordinating with the National PTA to amplify hashtags such as 27 #LearnOnTikTok and #TeachersOnTikTok. Ex. 475 at 6980, 6987; Ex. 476 at 8578. Internally, it credited 28 these efforts with boosting engagement under the #BackToSchool tag. Ex. 473. 85 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 98 of 235 1 Finally, TikTok leveraged its massive school-aged audience to attract advertisers during back-to- 2 school season. Once it secured a critical mass of student users, it promoted TikTok as the premier platform 3 for "back to school" campaigns—boasting an audience "full of school-aged children." Ex. 437 at 245:11- 4 246:2; Ex. 473. It told advertisers to expect "an influx of school supplies and clothing hauls from students 5 and parents alike," and "endless opportunities for brands to join in on the playground fun." Id. 6 TikTok closely tracked hashtags like #BackToSchool, touting a "massive jump in viewership" 7 from its "diverse audience of Gen-Z students, edu-Creators, and parents." Ex. 437 at 246:8-16; Ex. 473. 8 It knew children were using TikTok in school: "Between classes, during lunch, and after that final bell 9 rings, Gen-Z users... are pulling the squad together to dance, participate in trends, and create on TikTok." 10 Ex. 437 at 247:12-25. It even instructed advertisers how to appeal to this demographic—targeting 11 #collegeLife, #teacherLife, and #highSchool communities, promising, "you'll be the coolest kid in school 12 in no time." Ex. 437 at 248:20-249:7; Ex. 473. 13 (4) TikTok's knowledge of harms 14 Despite its heavily promoted platitudes about "digital safety," TikTok knew the design of its 15 platform posed serious risks—especially to minors. In a 2020 internal safety review, 16 , stated that social media is "inherently dangerous" and that minors are "at risk" on TikTok. 17 Ex. 530 at 4753; at 253:12-254:2. Through commissioned research, external partnerships, user 18 feedback, and internal studies, TikTok understood how its platform caused or worsened addiction, anxiety, 19 depression, sleep disruption, poor body image, and suicidality—and that it disrupted students and schools. 20 (a) TikTok's addiction problem was known, acknowledged, and profitable 21 Shortly after TikTok launched in the United States, TikTok knew its platform was habit-forming 22 and that it was fueling addictive and problematic use among children. As early as March 2019—barely 23 months after ByteDance rebranded Musical.ly as TikTok—executives received reports directly from 24 parents warning that "kids have become addicted to TikTok." Ex. 532 at 7593. One internal "Product 25 Experience Report" captured TikTok's attitude succinctly: while parent concerns about youth addiction 26 were "concerning for young minds," this phenomenon was viewed internally as "a good thing for the 27 platform." Id.; Ex. 635 at 360:4-19. Addiction, in other words, was not an unforeseen byproduct—it was 28 86 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 99 of 235 1 treated as a performance indicator. 2 The complaints multiplied year after year. Between 2018 and 2024, TikTok received countless 3 reports from its users and parents describing TikTok as "addictive," "like crystal meth," and "ruining" its 4 children's attention and sleep. Exs. 583, 584, 585, 586, 587, 588, 589, 590, 591. Users told TikTok they 5 were deleting their accounts because they were "getting addicted" or "trying to get over [its] TikTok 6 addiction." Id. Teachers and parents reported that students were "getting bad reports from teachers" 7 because of TikTok use. Ex. 589 at 2. 8 Internally, these warnings were heard but not acted upon. By 2020, TikTok's own internal 9 documents are replete with admissions like "addiction to technology is a ubiquitous problem that TikTok 10 and most other platforms deal with today." Ex. 534 at 1694; Ex. 632A at 228:8-21. The phrase "ubiquitous 11 problem" was not an excuse but an admission: TikTok recognized that its platform was driving the same 12 addiction pattern across millions of users, particularly school-aged users. 13 Even external consultants hired by TikTok confirmed this reality. , a 14 behavioral research group retained by TikTok, warned that the addictive experience was not incidental, it 15 was intentionally engineered. Ex. 581 at 0111. Yet TikTok continued to design and promote engagement- 16 maximizing features known to drive compulsive use, choosing profit over prevention. 17 The result was an established pattern of knowledge and inaction. Year-after-year, internal reports, 18 user feedback, and expert findings documented the same conclusion: TikTok was fostering addiction 19 among its users, particularly minors. TikTok knew it, quantified it, and continued to capitalize on it. 20 (b) TikTok's executives and employees struggled with addiction and its co-Defendants knew how addictive 21 TikTok was 22 Even TikTok's own executives and employees struggled to control their use of the platform. Their 23 personal experiences confirmed what TikTok's internal data had long shown: the product was engineered 24 for compulsion. , admitted under oath that he deleted the app and 25 installed a third-party blocker, the Freedom app, to curb his own use. at 277:1-278:9. was 26 not a teenager whose brain was still developing—he was a senior safety executive who needed outside 27 software to stop himself from using the very platform he was responsible for making safe. Id. at 281:24- 28 87 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 100 of 235 1 282:4. His experience laid bare the depth of TikTok's design: it could hook even the people charged with 2 protecting children from it. 3 was not alone. , admitted that 4 "TikTok is quite addicting." Ex. 676 at 3304. , reached the 5 same conclusion within two months of joining the company: "I'm addicted." Ex. 681 at 6089. These were 6 not casual users—they were the architects and guardians of the product, each personally experiencing the 7 dependency their platform produced in millions of children. 8 The recognition of TikTok's addictiveness extended beyond its own walls. Google—one of 9 TikTok's co-defendants and fiercest competitors—internally concluded that "TikTok can be extremely 10 addictive." Ex. 794 at slide 27. Similarly, , who testified in this action as Meta's corporate 11 designee, privately acknowledged to a colleague about TikTok: "I deleted the app because I was … 12 addicted." Ex. 1180 at 3628 (ellipsis in original). These are not outsider's critiques but acknowledgments 13 from rivals that fully understood the mechanics of engineered engagement. These assessments confirmed 14 what TikTok's insiders already knew: the platform was deliberately built to keep users hooked. 15 These admissions carry unique weight. were insiders with direct 16 access to TikTok's engagement metrics and safety systems. Their struggles were not coincidences—they 17 were symptoms of a product designed to capture attention and resist restraint. If senior executives— 18 experts in the very features driving addiction—could not limit their own use, the impact on children, 19 whose self-control and judgment are still developing, was not only foreseeable but inevitable. TikTok's 20 own leaders, not to mention its rivals, recognized what TikTok now pretends not to— addiction was not 21 a side effect of its design, but its success metric. 22 (c) TikTok intentionally engineered addiction 23 TikTok's own internal data and research confirmed what users, parents, and external experts had 24 long been saying: TikTok's platform was compulsive and habit-forming, producing patterns of excessive 25 use that harmed mental health and well-being. Far from being unaware, TikTok rigorously studied these 26 effects through surveys, focus groups, and internal analytics—and repeatedly documented addiction- 27 related concerns in its internal files. To date, TikTok has hidden those files from the public. 28 88 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 101 of 235 1 As early as 2020, TikTok's internal surveys showed that a significant share of users actively tried 2 to reduce or quit using TikTok because they were spending "too much time" on the app. Ex. 535 at 2294. 3 In these surveys, "fear of getting addicted" ranked among the leading reasons users deleted the app 4 altogether. Ex. 536 at 3102. Users repeatedly described feeling hooked and unable to stop scrolling, 5 expressing guilt and frustration over the hours lost each day on the app. Ex. 537 at 3040. These findings 6 were strikingly consistent across demographic groups, but particularly pronounced among youth, 7 underscoring that the problem was not isolated—it was systemic. 8 A 2021 internal analysis by TikTok's Trust & Safety team went even further. It explicitly 9 acknowledged that "our users' biggest usage deterrent is that they think the platform is addictive," and 10 that "compulsive usage correlates with a slew of negative mental effects, like loss of analytical skills, 11 memory formation, contextual thinking, conversational depth, empathy and increased anxiety." Ex. 537 12 at 3040-3041; Ex. 632A at 259:1-7. The same document warned that users were caught in a loop of 13 automatic engagement, which deprioritized self-control. Ex. 537 at 3044. The internal analysis concluded 14 candidly that "compulsive usage on TikTok is rampant and our users need better tools to understand its 15 usage [and] manage it effectively." Ex. 537 at 3041; see id. at 3040 ("users' biggest usage deterrent is that 16 they think the platform is addictive."); Ex. 632A at 257:13-25, 260:8-19. 17 TikTok's internal documents also make clear that the platform's addictive pull was engineered, 18 not accidental. TikTok's behavioral research confirmed that the very features driving engagement were 19 producing the cognitive dependence users described as "addiction." Ex. 537 at 3040-3041. Researchers 20 noted that the platform's variable reward structure conditioned users to seek intermittent dopamine 21 rewards through endless swiping, mirroring behavioral reinforcement patterns found in gambling. Ex. 508 22 at 179:20-25, 181:15-182:24. Internal documents described the "habit moment" —the point at which use 23 transitions from voluntary to automatic—which occurred after approximately 260 videos in the first week. 24 Ex. 538 at 29; Ex. 433 at 159:6-16. , 25 confirmed that the platform intentionally included a "slot-machine effect" that drove repeated engagement 26 and stripped users of meaningful control. at 82:23-83:8; Ex. 541 at 9176; Ex. 542 at 1765. This 27 recognition of the "slot machine effect" appeared frequently across TikTok's own internal documents. Id.; 28 see also Ex. 556 at 1514. 89 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 102 of 235 1 These findings did not come from isolated teams—they reflected company-wide recognition of 2 the problem. Product designers, data scientists, and Trust & Safety personnel all contributed to the 3 analyses. Further, TikTok saw that these compulsive patterns carried measurable harmful psychological 4 consequences. Internal data linked excessive TikTok use to heightened anxiety, sleep deprivation, and 5 decreased academic focus—effects TikTok categorized as "negative mental effects." Ex. 537 at 3040- 6 3041. and others acknowledged that TikTok's optimization for engagement produced 7 rampant compulsive usage, correlating with anxiety, sleep disruption, and loss of concentration in school. 8 at 229:7-12, 260:8-19. TikTok's own data confirmed that school-aged users were among the 9 most affected. Internal analyses acknowledged that students were active on the platform late at night, 10 arriving at school sleep-deprived, and checking notifications during class—behaviors the company 11 recognized as foreseeable outcomes of its engagement strategy. Ex. 413B at 202:14-21; Ex. 643 at 12 8755. 13 TikTok could have addressed the fundamental issue: Its platform design was creating addiction 14 and mental health risks. Instead, TikTok continued to optimize for "time spent" and "retention," not user 15 health Ex. 556 at 1514. Internal documents make clear that TikTok's financial goals are directly tied to its 16 uniquely "powerful" and "coercive" nature. Id. TikTok's internal dashboards 17 18 . Ex. 632A at 257:13-25. 19 . Ex. 556 at 1514. 20 In short, TikTok's own research (still not public) confirmed that addiction was not an unintended 21 byproduct of the platform but was an integral, measurable outcome. The very executives and employees 22 responsible for mitigating harm could not themselves disengage, underscoring the depth of the problem 23 and the deliberate nature of the addiction. Rather than putting up guardrails to protect school-aged 24 children, TikTok built guardrails only to protect its revenue—accepting widespread addiction as the price 25 of maximizing engagement and, by extension, profit. Rather than mitigating these harms, TikTok 26 reinforced them, because every additional minute on the platform translated into more advertising revenue. 27 Rather than warning users, parents, and educators about the problem, TikTok buried its research. 28 90 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 103 of 235 1 (d) TikTok knew its platform was harming users' mental health, sleep, and schools - but chose profit over 2 prevention 3 TikTok's internal documents leave no doubt: it knew its platform was harming users' mental 4 health, disrupting sleep, and undermining the classroom environment. Those harms were not theoretical. 5 Internal studies, surveys, user feedback, and external research commissioned by TikTok repeatedly 6 identified that excessive TikTok use is correlated with anxiety, depression, and diminished self-esteem. 7 See, e.g., Ex. 594 at 6419 (Trust & Safety team conceding that "compulsive usage correlates with a slew 8 of negative mental effects, including increased anxiety," and that the platform "exacerbate[es] preexisting 9 mental health conditions such as depression and anxiety."); Ex. 632A at 259:1-7. 10 By 2020 at the latest, internal safety reviews warned that social media is "inherently dangerous" 11 and that minors were "at risk" on TikTok. Ex. 530 at 4753; Ex. 402A at 253:12-254:2. An internal 2021 12 report captured the problem starkly: users described feeling "sucked in" and reported a range of negative 13 emotions after extended use, including "guilt, frustration, irritation, disappointment, and anxiety." Ex. 592 14 at 2102. TikTok conducted numerous focus groups, which confirmed that teens "acknowledge that they 15 sometimes lack agency on TikTok," and that this loss of control was directly tied to stress and anxiety. 16 Ex. 593 at 9673. 17 TikTok also tracked and reviewed external studies that reached the same conclusion. Internal 18 memoranda cited peer-reviewed research showing that minors who spend more than three hours a day on 19 social media double their risk of "poor mental health outcomes, including symptoms of depression and 20 anxiety." Ex. 595 at 0819; Ex. 596 at 0680. Another study, included in TikTok's Minor Safety Strategy 21 Paper, linked six or more hours of daily screen use to heightened depression risk. Ex. 485 at 0816. Yet 22 TikTok's own data confirmed that its users—especially minors—were routinely exceeding that threshold. 23 Nearly users spent more than on TikTok, prompting an internal 24 acknowledgment that a "stronger approach" was needed to curb "harmful screen time." Id. 25 , personally recognized that "depression and addiction" were "conditions 26 often linked to the use of apps and social networks," acknowledging that these risks gave parents "great 27 apprehension." Ex. 597. 28 91 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 104 of 235 1 Despite this, TikTok took no meaningful steps to mitigate harm. Instead, it continued to privately 2 optimize engagement (and revenue) by deepening the very design patterns its data tied to mental distress. 3 And it concealed its troubling findings, instead publicly promoting "digital-wellbeing." 4 TikTok's awareness extended beyond addiction and anxiety. Internal analyses and expert feedback 5 directly linked TikTok use to self-harm and suicide. In one 2021 report, TikTok's safety researchers 6 concluded that the risk of users being driven into filter bubbles involving suicide and self-harm material 7 was "HIGH." Ex. 633 at 374:21-376:13; Ex. 517 at 8485. TikTok's internal risk assessments even asked 8 whether the platform was "seeding suicidal and self-harm ideation in young people who had never 9 previously contemplated it." Ex. 599 at slide 42. These internal findings mirrored public health warnings 10 that excessive algorithmic exposure could normalize self-destructive behavior—warnings TikTok 11 ignored. Ex. 1184. 12 TikTok also received explicit notice from experts about the dangers of the platform promoting 13 eating disorders. , a public health specialist on adolescent eating disorders, advised TikTok 14 to add warning labels regarding that risk. Ex. 492 at 60:25-70:21; Ex. 659 at 3337; see also Ex. 667 at 2. 15 TikTok refused. Instead, it contracted with and paid Eugenia Cooney, an influencer "widely known in the 16 Eating Disorder Community" for severe anorexia, effectively turning her into a spokesperson for 17 disordered behavior. Ex. 492 at 57:11-22, 99:5-24; Ex. 688; see also Ex. 492 at 62:21-66:19; Ex. 659 at 18 3333, 3335-37; Ex. 672B at 343:6-370:22. 19 TikTok also knew its platform interfered with sleep and learning—two pillars of child 20 development. TikTok's leadership understood that teenagers require 8–10 hours of sleep per night to 21 function properly. Ex. 402A at 134:10-135:8; Ex. 604 at 785. Yet TikTok's internal analyses 22 acknowledged that "social media use at night" is associated with "delayed and inadequate sleep" and that 23 "users are active on TikTok when they should be sleeping." Ex. 632A at 278:10-283:24; Ex. 485 at 0817; 24 Ex. 601 at 9294. Its own data showed that of users aged 13–15 and of those aged 16–17 were 25 active between midnight and 5 a.m. Ex. 632A at 282:4–285:17. The Trust & Safety team conceded that 26 "the compulsion to react to notifications at night can result in lower-quality sleep," and that this problem 27 "impacts our minor users significantly." Ex. 602 at 1842; Ex. 603 at 8036. Still, TikTok sent notifications 28 to users during the school day and in some cases, up until midnight, which could interfere with sleep. Ex. 92 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 105 of 235 1 540 at 1075; Ex. 603 at 8036 ( received notifications at "midnight or beyond"). Far 2 from restricting nighttime notifications, or warning young users about their risks, TikTok knowingly 3 disrupted users' rest cycles—the very harm its executives had internally acknowledged. 4 The same disregard extended to the school environment. Internal discussions admitted that sending 5 notifications during school hours (9 a.m. to 3 p.m.) was "disruptive" and risked "interfering with study at 6 school." Ex. 413B at 202:14-21, 219:25-20:5; Ex. 643 at 8755; Ex. 606 at 2860. Nevertheless, internal 7 analyses—never made public—showed " ," 8 proving that TikTok's systems encouraged in-class engagement among minors. Ex. 413B at 344:5-10; Ex. 9 649 at 26-27. 10 When schools and parent organizations raised concerns, TikTok's response was denial—not 11 reform. Instead of addressing the problem, it instructed the National PTA, one of its paid partners, to send 12 letters claiming TikTok did "not currently see evidence" of school-related challenges—even as TikTok 13 internally tracked over 200,000 videos about the "Devious Licks" challenge disrupting classrooms 14 nationwide. Ex. 444A at 174:23-177:3, 189:12-193:18; Ex. 615 at 4015; Ex. 616 at 5062. 15 (5) TikTok's failure to exercise reasonable care 16 TikTok knew its platform was harming children and disrupting schools but chose to do nothing. 17 Internal documents showed students losing sleep, experiencing anxiety, and using TikTok compulsively 18 during class. Rather than address these risks, TikTok concealed the dangers, ignored available safeguards, 19 and prioritized engagement over safety—even as it implemented stronger protections for minors in the 20 Chinese version of TikTok (Douyin) that were never introduced in the U.S. Its intentional and conscious 21 disregard for foreseeable harm has caused widespread classroom distraction, behavioral problems, and 22 drained school resources. As discussed below, this failure to exercise reasonable care—through weak age 23 verification, ineffective parental controls, and other safety lapses—has created a public nuisance that 24 continues to harm and burden the Districts. 25 (a) Age Verification 26 TikTok's own policies restrict use of the platform to individuals aged 13 and older, meaning that 27 elementary and most middle school students are supposed to be ineligible. Ex. 400 at 19-20. Nonetheless, 28 it chose not to implement effective age verification measures to prevent these children—including those 93 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 106 of 235 1 using the app at school—from accessing the platform. Ex. 404 at 122:22-123:22; Ex. 403 at 8049 (for age 2 assurance, TikTok relies solely "on an age gate (self-declaration) in all markets"). 3 TikTok's documents confirm that it has long been aware of underage use. Internal estimates 4 showed that at least of TikTok's users were under the age of 13, representing more than 5 underage accounts in the United States alone. Ex. 418 at 45:6-46:3; Ex. 419 at 2064; Ex. 683A at 110:13- 6 112:14; Ex. 420 at 7491. Other internal reports document an 11-year-old user in 2018, a 7-year-old in 7 2019, a 10-year-old in 2020, an 11-year-old in 2021, an 11-year-old in 2022, and an 11-year-old in 2023. 8 Ex. 672A at 192:18-193:22, 271:20-272:16, 196:7-197:6, 202:21-204:11, 209:25-211:8, 211:14-213:1; 9 Ex. 650 at 2; Ex. 655 at 2; Ex. 651 at 2; Ex. 652 at 2; Ex. 653 at 2; Ex. 654 at 2. 10 Although direct messaging was disabled for users under 16, Ex. 402A at 170:5-171:1, TikTok has 11 never verified users' ages—despite acknowledging that age assurance is "the foundation of every Minor 12 Safety and Privacy strategy at TikTok." Ex. 404 at 122:22-123:16, 262:7-264:11; Ex. 401 at 7386; Ex. 13 403 at 8049 (relying instead on the age gate). 14 Until February 2019, TikTok had no age gate at all. When TikTok finally introduced one, it merely 15 asked users to self-report their birthdates. Ex. 400 at 18. TikTok employees acknowledged that this 16 measure was ineffective and easily evaded. , admitted 17 that "all you have to do is look at the age gate and understand that it's easy to put in whatever." 18 at 222:8–223:2. Another , was blunter: "[W]e are idiots to think 19 a kid these days will not cheat the system." at 259:23-263:18; Ex. 421 at 2641. 20 Despite knowing these flaws, TikTok never implemented available tools—such as ID checks, 21 biometric verification, or credit card confirmation—to verify user age at sign-up. Ex. 409 at slide 19; Ex. 22 404 at 122:22-123:16; Ex. 403 at 8049 (relying solely on the age gate). 23 TikTok understood that enforcing its own age restrictions would come at a huge financial cost. 24 When , considered whether TikTok should "remove all U13 25 [under-13] users," he argued against doing so, citing concerns about losing too many users—and, with 26 them, too much revenue. at 186:1-12, 189:18-191:7, 192:21-25; Ex. 478. 27 The failure to adopt stronger age safeguards is not a matter of speculation or a lack of know-how. 28 Nearly two years before TikTok's U.S. launch, ByteDance introduced the Chinese version of TikTok, 94 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 107 of 235 1 Douyin, in September 2016. Ex. 635 at 94:13-21; Ex. 549 at 2915. The U.S. version of TikTok was built 2 on the same underlying algorithm and core features as the Chinese version (Douyin). Ex. 402B at 480:9- 3 11; Ex. 632B at 450:7-10. The two platforms share nearly identical functionalities—including short-form 4 videos, the "For You" feed, infinite scroll, and push notifications. Ex. 632B at 450:7-17. 5 A key difference between the Chinese TikTok (Douyin) and the U.S. version lies in age 6 verification and authentication. Chinese TikTok (Douyin) employs an "Age Model" (also referred to as 7 "Age Assurance") to actively verify user ages and trigger interventions that ensure accurate identification 8 of minors for safety features. Ex. 632B at 452:25-453:3; Ex. 496 at 1116 ("Douyin's Stance . . . Important 9 to note they use their age model to deliver interventions"). Chinese TikTok (Douyin) also requires real- 10 name authentication for minor users. Ex. 632B at 467:23-468:6, 572:20-573:12; Ex. 552 at 8104. 11 The U.S. version, by contrast, does not use the Age Model. Instead, it relies solely on self-reported 12 birthdates through an "Age Gate," a system that depends on user honesty rather than verification. This 13 difference has significant consequences: when minors create accounts using false ages, TikTok's minor 14 safety features are never activated. Ex. 632B at 468:19-469:21. 15 TikTok fully understood the importance of accurate age determination through the Age Model 16 system. A July 2022 internal document, titled "Minor Safety 2022 H2 Roadmap," acknowledged that "[i]n 17 order to protect minors, we need to know who on our platform is a minor"—a statement with which 18 , expressly agreed. at 471:8- 19 11; Ex. 555 at 4857. In the same document, TikTok admitted that the existing "Age Gate" [in the U.S.] 20 was ineffective because users could easily falsify their birthdates: "relying on age gate turns out to be less 21 effective because there is no way for us to ensure that users are accurately entering their true date of birth." 22 Id. at 4862; Ex. 632B at 472:19-473:4. Despite recognizing these flaws, TikTok declined to implement 23 the Age Model used in the Chinese TikTok (Douyin) and continued to rely on the defective Age Gate 24 system in the United States. Id. 25 Rather than strengthening its age gate to prevent underage access, TikTok made it easier to bypass. 26 It introduced "Guest Mode," allowing users to view TikTok videos without creating or signing into an 27 account. Ex. 406 at 3121-3122. TikTok's business team projected that 28 95 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 108 of 235 1 ." Ex. 407 at 8119. Likewise, TikTok permitted users to avoid the age gate entirely by 2 registering through Facebook or Google accounts. Ex. 697 at 156:21-158:21; Ex. 408 at 6536. 3 TikTok had the capability to identify and remove underage users through its age inference model 4 but chose not to use it for that purpose in the U.S. For every user, TikTok maintains both a self-reported 5 age and an age inferred by its machine-learning model. Ex. 410 at 23-24. It knew its inferred age data was 6 more accurate and used it when it served its business interests—such as recommending videos or targeting 7 ads based on a user's actual age. Ex. 427B at 617:21-618:24, 621:21-625:8; Ex. 416 at 3077-3078; Ex. 8 417 at 21-22. Yet in the United States, TikTok deliberately declined to apply that same model to identify 9 and remove suspected underage accounts. Ex. 404 at 124:9-125:17. 10 TikTok went so far as to mislead regulators about its ability to detect underage users. When asked 11 whether it had any age models capable of identifying or removing such accounts, the company falsely 12 claimed it did not—despite possessing exactly that capability. Ex. 427B at 629:13-632:22; Ex. 480 at 13 2929A-2930A. Internally, employees acknowledged that if EU or U.S. regulators learned the truth, 14 TikTok might be forced "to delete a large volume of u13 [under-13] users," resulting in an estimated 15 loss of its user base. Ex. 427B at 636:18-24; Ex. 480 at 2930A. 16 TikTok had a clear incentive not to identify children under 13: knowing meant removing them, 17 and removing them meant losing revenue. To preserve its underage user base, the company made decisions 18 to feign ignorance of its users' true ages. In June 2018, as Musical.ly prepared to rebrand as TikTok, 19 , a member of the rebranding team, proposed surveying users to learn their age. Ex. 433 at 87:8-15, 20 88:2-89:1; Ex. 568 at 9300. He suggested using education level as a proxy, including whether users were 21 in middle or high school. Id. Musical.ly co-founder and immediately 22 rejected the idea, noting that including "elementary school" as an indicator would be "definitely violating 23 COPPA," and admitting: "[i]f [the FTC] knows about this survey, they might want us to submit the results, 24 and I am not sure if it's good result." at 89:21-90:2, 93:2-15; Ex. 568 at 9301. Another executive, 25 , warned, "we must avoid any age demographic questions. Doing school level proxy 26 would still be considered age demographics. We take the position in various cases and regulatory matters 27 that we do not collect such information." at 95:17-96:22; Ex. 568 at 9301. As a result, 28 "made sure none of the age-related questions will appear." Id. 96 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 109 of 235 1 TikTok then delayed meaningful action for years rather than remove underage users from its U.S. 2 platform. Its Standard Operating Procedures instructed employees to flag accounts only if a user appeared 3 to be "9 years old" and to ignore parental requests to delete teen accounts. Ex. 697 at 173:2-14; Ex. 481 4 at 2015. The inaction was not due to technical limitations—it was a choice. When political scrutiny and 5 the threat of a U.S. ban intensified in 2024, TikTok deleted more than 21 million underage accounts in 6 just three months. Ex. 400 at 20. By then, however, the damage to students and school districts was done. 7 TikTok's willful ignorance extended beyond data collection. TikTok concealed from the public— 8 and from school districts—that children routinely falsified their ages to bypass the platform's restrictions 9 or could evade the age gate entirely by registering through other platforms like Google. Ex. 444A at 10 262:25-263:4. It also failed to alert schools that they might need to devote resources to curbing student 11 use during instructional hours. Ex. 444B at 365:19-366:6. By hiding the scope of underage use and 12 withholding meaningful safeguards, TikTok deprived schools and parents of the information necessary to 13 protect children—choosing growth and profit over even minimal responsibility for their safety. 14 (b) Parental Controls 15 TikTok promoted Family Pairing as one of its signature child safety tools but internally 16 acknowledged that it was deeply flawed. From the outset, it knew design barriers discouraged parents 17 from setting up Family Pairing at all. Ex. 405A at 107:4-10; Ex. 490 at 7039. Even when parents did 18 enroll, TikTok allowed children to unlink their accounts freely—rendering the feature, in its employees' 19 words, "kinda useless." Ex. 632B at 634:10-635:22; Ex. 491 at 6727; Ex. 405A at 48:19-22. TikTok also 20 knew minors were falsely registering as "parents" to bypass restrictions because the system lacked age 21 verification. Ex. 484 at 2445. Parents also received no notice outside the platform itself when their child 22 unlinked from Family Pairing, depriving them of any opportunity to address it. Ex. 405A at 69:18-25. 23 Internally, TikTok admitted that Family Pairing was rushed and poorly maintained. The feature 24 was pushed through "speedy development" driven by "urgent action in response to regulators," not user 25 safety, and its rollout automatically reset children's privacy settings to "public" when accounts were linked 26 or unlinked. Ex. 405A at 124:25-126:3, 142:9-17, 143:16–144:2; Ex. 493 at 8936, 8939-40; Ex. 494 at 27 0132. As , the TikTok leader responsible for Family Pairing, put it, "Family Pairing 28 is where all good product design goes to die[.]" at 125:12-18; Ex. 493 at 8939. Despite knowing 97 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 110 of 235 1 these flaws, TikTok never told schools or parents that teens could unlink their accounts or bypass the 2 controls. Ex. 444A at 258:17-23, 259:21-260:4; Ex. 444B at 352:12-21, 353:10-17. Nor did it disclose 3 internal test results showing screen-time management tools were ineffective. Ex. 405A at 199:24-200:22. 4 In early 2023, TikTok rushed out new "parental control" features under Project M ("M" stands for 5 "Minor"), setting the stage for CEO Shou Chew's polished sales pitch before the United States Congress. 6 Ex. 405B at 318:24-319:21, 333:6-12; Ex. 479 at 4627-28. The Trust & Safety team was instructed to 7 release the tools "by March 23rd since that's when Shou is testifying in front of [C]ongress." Id. Internally, 8 Family Pairing was not described as a safety tool or feature but rather as a "PR defensive weapon" (Ex. 9 489 at 2510) and "a very, very loud way to build awareness around Family Pairing" for regulatory 10 appearances. Ex. 482 at 1893; Ex. 405B at 316:21-317:11. 11 TikTok's remarked that "Family 12 Pairing finally has its moment in the light." at 323:2-324:5; Ex. 1187 at 4868. His relief 13 underscored how long the company had postponed safety improvements. The renewed push came only 14 after "executive sponsorship" from —reflecting that TikTok acts on safety measures not 15 out of concern for children but when protecting its brand demands it. at 324:7-22, 330:7-20; Ex. 16 1187 at 4869 ("it's coming down from and so we've got momentum"); Ex. 567 at 7300 17 ("It's nothing new, sadly. Just all the things we talked about last year.") (emphasis added). 18 ByteDance and TikTok's leadership also made clear that new features could not meaningfully 19 reduce engagement: screen-time tools were permitted to "hurt" metrics by no more than for extreme 20 users and minors. Ex. 632A at 278:10-23, 312:4-314:2; Ex. 485 at 0820. When usage dropped, executives 21 questioned "by how much," noting that "[e]ven a few minutes fewer means fewer ads and the impact on 22 revenue at scale is significant." Ex. 405A at 208:20-209:3; Ex. 486 at 7110. TikTok therefore rejected a 23 proposal to adopt a screen-time limit that users could not bypass. Ex. 483 at 5464 (rejecting a hard limit). 24 Internal tests confirmed that these so-called safety tools had no meaningful effect: TikTok found no 25 "negative impact" on minors' average session duration or on key metrics such as daily active users and 26 retention. Ex. 405A at 199:6-7, 199:24-200:5; Ex. 487 at 7228, 7232, 7234; Ex. 488 at 7711. But it 27 withheld these results, admitting "it's not practice to disclose" such data publicly. Ex. 405A at 200:6-22. 28 98 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 111 of 235 1 In short, TikTok's approach to parental controls prioritized optics—and not losing users—over 2 safety. Although Family Pairing was promoted publicly, including by the CEO in testimony before 3 Congress, as proof of TikTok's dedication to protecting minors, internal documents reveal it was intended 4 primarily as a public relations strategy rather than a meaningful safeguard. The Defendants knew children 5 could easily bypass age and usage restrictions and that its so-called screen-time limits made no meaningful 6 difference. Acknowledging these failures, however, risked reputational and financial harm. Rather than 7 addressing the problem, TikTok chose to protect its image and profits—demonstrating a blatant disregard 8 for the families and schools it claimed to serve. 9 In promoting Family Pairing as one of its flagship child-safety tools, TikTok publicly proclaimed 10 its noble commitment "to giving parents insight into, and control over," their child's account. Ex. 699 at 11 3. Yet those assurances are irreconcilable with its internal policies, which deliberately strip parents of the 12 very authority the company claimed to provide. TikTok's own standard operating procedures make this 13 contradiction unmistakable. Employees tasked with communicating with the public are expressly 14 instructed to deny parental requests to delete a child's account if the child is thirteen or older. Ex. 481 at 15 2015 ("If a parent reaches out to delete their child's account, but the child is age 13–17, we cannot delete 16 the account."); Ex. 698 at 2079; Ex. 697 at 87:8-88:12. Worse still, if a parent persists—renewing the 17 deletion request—employees are directed to ignore the parent entirely. Ex. 698 at 2079 ("If the parent 18 writes back again and gives pushback on the same request of deletion, do not respond.") (emphasis in 19 original); Ex. 481 at 2015; Ex. 697 at 89:17-90:9. 20 TikTok's disregard for child safety extends beyond families to schools and educators. Reports 21 from teachers or administrators concerning users aged 13–17 are categorically disregarded. Ex. 698 at 22 2077 (flow chart showing that when a third party—such as a teacher—reports a user aged thirteen or older, 23 no action is taken). Through these policies, TikTok has positioned itself as the ultimate arbiter of what is 24 "best" for teens—overriding the judgment of parents, educators, and others who actually know and care 25 for the child. TikTok's private conduct stands in stark contrast to its public promises of parental 26 empowerment and child safety. 27 (c) Screen Time Management 28 TikTok did not implement screen-time management ("STM") tools to curb use, but to counter 99 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 112 of 235 1 growing criticism and reassure regulators the platform was safe. Internal research confirmed that many 2 potential users avoided TikTok because they believed "the platform is addictive." Ex. 632A at 257:21-25; 3 Ex. 495 at 8276. That insight—rather than any genuine concern for user well-being—became the driving 4 force behind the development of TikTok's screen-time features. Id. 5 Internal TikTok documents make clear that the objective of these tools was not behavioral change 6 but managing external perception and protecting the brand and revenue. A high-level product document 7 explicitly stated, "Our goal is not to reduce time spent" on the app, but to influence how users viewed 8 TikTok to increase "DAU [daily active users] and retention." Ex. 485 at 0821. Another described the 9 screen-time controls as "opportunities to improve stakeholders' perception of TikTok as safe for 10 teenagers." Ex. 496 at 1111. 11 Consistent with that purpose, internal records confirm that TikTok imposed strict limitations to 12 ensure screen-time features would not materially affect engagement or revenue. See, e.g., Ex. 485 at 0820. 13 For all screen-time management projects, , capped any acceptable 14 reduction in user time at no more than . Id. ("accept up to decrease"). Internal performance 15 analyses further showed that reducing use by merely would have exceeded that 16 threshold. at 345:20-23. Reflecting those constraints, internal Trust & Safety planning materials 17 show that the team pursued only "moderate" screen-time management initiatives specifically designed to 18 preserve engagement rather than meaningfully curtail it. Ex. 496 at 1112. 19 When TikTok introduced features allowing users to track their time and set screen-time breaks, it 20 made those tools entirely optional. Ex. 497 at 2841. TikTok's 21 , acknowledged internally that "anything opt in" receives " [.]" Ex. 498 22 at 1898; at 181:08-25. Pre-launch testing confirmed as much: only of users were expected 23 to use the tools, resulting in less than a reduction in time spent on the platform—and no reduction 24 at all for teens—because "minors do not have [the] executive mental function to control their screen 25 time[.]" Ex. 497 at 2841. Nonetheless, TikTok released them anyway—and buried the features behind a 26 "hidden series of menus." Ex. 632B at 593:18–21; Ex. 495 at 8274. 27 TikTok also introduced "Addiction"—later renamed "Take a Break"—videos to encourage users 28 to log off after extended use. Ex. 499 at 1549. However, those videos were deliberately designed to be 100 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 113 of 235 1 skippable, allowing users to bypass them with "very low friction." Id. TikTok's own data showed that 2 over of users simply swiped past the prompts, and many left comments mocking the feature's 3 effectiveness. Id. at 1549-50. 4 By late 2022, facing "intensifying criticism at the highest levels of US and EU politicians about 5 addiction-related harm among teens on TikTok," TikTok's Government Relations team urged the 6 company to release a new screen-time feature "quickly in the hope of reducing or redirecting some of this 7 pressure." Ex. 632A at 362:17-363:4, 364:17-23; Ex. 500 at 5806. Although the Trust & Safety team 8 proposed a genuine "hard cap" on teen usage, Government Relations instead advocated for a default limit 9 that "[t]eens can turn off[.]" Ex. 500 at 5805, 5810. The result was a 60-minute default time limit for 10 teens—easily bypassed by entering a four-digit PIN code. Ex. 632B at 537:3-16. 11 TikTok possessed both the capability and the user feedback necessary to design meaningful screen- 12 time controls but chose not to implement them. In company-run focus groups, half of the teen participants 13 reported that they wanted time-management tools that would prevent them from "simply ignoring [limits] 14 when a limit is reached," such as by "shutting off the app[.]" Ex. 501 at 6136. TikTok, however, only 15 implemented such features in China. 16 The contrast between TikTok in China (Douyin) and the U.S. platform is stark. TikTok 17 implemented a digital well-being system explicitly focused on addiction prevention on Chinese TikTok 18 (Douyin), aiming to help teenagers "truly and effectively control" their time and "reduce the proportion 19 of users who use Douyin for a long time." Ex. 552 at 8102-8103; Ex. 632B at 459:15-463:8. By contrast, 20 in the U.S., TikTok emphasized user agency rather than addiction mitigation. As one internal document 21 acknowledged, the "[k]ey distinction in our work is Douyin's focus on addiction, while ours is on agency," 22 and TikTok's stated goal in the U.S. was 23 . Ex. 552 at 8102; Ex. 485 at 0821. 24 As early as 2021, TikTok implemented strict protections for verified users under 14 on the Chinese 25 TikTok (Douyin), including a mandatory youth mode that imposed a hard cap of 40 minutes per day and 26 prohibited use between 10 p.m. and 6 a.m. Ex. 632B at 473:15-22, 482:10-483:10, 487:11-19. By contrast, 27 TikTok did not introduce its 60-minute daily screen-time limit for minors in the U.S. until 2023—and 28 unlike the Chinese version, in the U.S. that limit was easily bypassed, with no nighttime restrictions of 101 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 114 of 235 1 any kind. Ex. 632B at 411:23-414:11. Internal documents show that TikTok rejected stronger options out 2 of concern for reducing "stay duration." Id. Unsurprisingly, its own data later confirmed that excessive 3 use among minors remained widespread: minors averaged per day on the platform, and were 4 "heavy users" spending more than six hours daily. Ex. 632B at 440:22-441:8; Ex. 544 at 9949. 5 TikTok was fully aware that minors were using the platform late at night and that excessive 6 nighttime screen use contributed to inadequate sleep and related harms. Ex. 485 at 0816. Internal data 7 from 2022 showed that more than of minors were active on the platform between 12 a.m. and 5 a.m. 8 Id. at 0817. Despite this knowledge, it never implemented a hard cap or any effective restriction on 9 nighttime use by minors. Ex. 632B at 488:17-489:13. 10 TikTok also implemented other meaningful safeguards on the Chinese TikTok (Douyin) to limit 11 excessive use among minors—safeguards that were not implemented in the U.S. Beginning in late 2021, 12 TikTok introduced, on the Chinese TikTok (Douyin), blocking layers and forced rest periods that could 13 not be bypassed by users, along with mandatory breaks after extended use. Ex. 632B at 456:1-15 14 (discussing Douyin's reminder system that "can't be bypassed by the user"), 473:23-476:17 (describing 15 Douyin's "forced rest" periods for users ages 14 to 18 and adults that "can't be scratched away"); Ex. 496 16 at 1116-1117 (comparing "Douyin's Stance"); Ex. 552 at 8104 (discussing blocking layers); Ex. 554 at 1 17 (an example of Douyin's reminders). They also added a five-second mandatory pause between videos on 18 the Chinese version for minors—again, a feature users could not skip. Ex. 557; Ex. 580 at 5402. In 19 contrast, TikTok maintained a "user agency" approach in the U.S., relying on self-control rather than 20 mandatory limits, despite knowing that Douyin's stricter controls significantly reduced excessive use. Id. 21 Despite being owned and operated by the same parent company, ByteDance, the Chinese TikTok 22 (Douyin) employs far more restrictive and effective screen-time management measures than the U.S. 23 version, as reflected in Figure 1 below. Technology experts have noted this disparity, describing China's 24 Douyin as a "spinach" version of the app designed to promote healthy use, while the U.S. version is the 25 "opium" version—engineered for addiction and profit. See Ex. 506 at 2-3 (quoting Tristan Harris from 26 Oct. 2022). 27 28 102 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 115 of 235 1 2 3 4 5 6 7 8 9 10 11 Figure 1. 12 TikTok knew that its screen-time tools in the U.S. were ineffective—both in theory and in 13 practice—yet failed to disclose this information to parents, students, or schools. Ex. 444B at 484:14- 14 490:15. TikTok also possessed both the knowledge and technical capability to build a safer experience for 15 minors in the United States (as was done in China) but chose not to do so. Ex. 402B at 486:3–7, 486:10; 16 Ex. 632B at 581:6-13, 627:3-7 (confirming that nothing prevented ByteDance from launching TikTok 17 with the same protective features implemented on the Chinese TikTok (Douyin)). 18 In sum, TikTok's approach to screen-time management reflects a deliberate pattern of prioritizing 19 growth and public perception over safety. TikTok knew its tools were ineffective, declined to adopt proven 20 safeguards already implemented on the Chinese TikTok (Douyin), and concealed these deficiencies from 21 parents, schools, and regulators. By choosing to protect its business metrics rather than its youngest users, 22 TikTok failed to exercise reasonable care toward the very minors it publicly claimed to protect. 23 (d) "The Ugly Side of Beauty Filters" 24 From the outset, TikTok knew its beauty filters could harm users' mental health and body image, 25 particularly in adolescents—yet chose to proceed anyway. As early as 2018, Musical.ly (later acquired by 26 ByteDance and rebranded as TikTok) faced complaints that its "skinny filter" promoted negative body 27 image, but TikTok deemed those risks "acceptable." Ex. 617 at 0909; Ex. 618. This decision reflected a 28 broader corporate pattern: prioritizing growth and engagement over user well-being. From day one, beauty 103 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 116 of 235 1 filters were embedded into the platform. Ex. 619 at 44; Ex. 632A at 114:10–25. 2 Despite growing public concern, TikTok ignored repeated warnings that its filters were damaging 3 users' self-esteem, particularly among school-aged users and girls. Media reports throughout 2020 and 4 2021 raised alarms. Ex. 632A at 184:24-185:6, 196:10-23, 197:12-201:7, 201:14-17, 203:13-205:15, 5 206:6-207:16, 208:4-210:3; Ex. 620 at 5106. Users themselves pleaded with TikTok to remove or label 6 filters that "trigger people's mental illnesses, body dysmorphia or dysphoria." Ex. 672A at 231:1–3, 232:5- 7 23, 234:9-15, 235:17-238:9, 240:20-242:24, 243:8-21, 245:11-247:7, 248:21-249:4, 249:12-23; Exs. 627- 8 631 (user complaints about the presence of filters). 9 Internally, TikTok's Trust & Safety team analyzed and wrote a memo about "The Ugly Side of 10 Beauty Filters," documenting harms such as "Snapchat dysmorphia," the normalization of cosmetic 11 surgery, and the disproportionate impact on children. Ex. 622 at 4863. A 2022 internal report warned that 12 filters emphasizing thinness, smoothness, or youth could severely damage young users' self-image— 13 specifically identifying "thin," "beautifying," and "age" filters as particularly harmful. Ex. 632A at 14 191:20-195:6, 195:14-196:4; Ex. 620 at 5102-04. 15 Despite this knowledge, TikTok did not curtail the use of beauty filters—it expanded it. In 2023, 16 it released a new wave of powerful, AI-driven filters that epitomized the dangers its safety team had 17 warned against. In February, it launched the "Teenage Look" filter, which artificially makes users appear 18 younger, and the "Bold Glamour" filter, which overlays a digital mask of makeup and "perfected" facial 19 features. Ex. 632A at 171:6–20, 172:14–173:16, 174:10–13; Ex. 621. These filters are depicted in Figure 20 2. See Ex. 632A at 179:22-180:5; Ex. 694. 21 Then, in August of the same year, TikTok went further, introducing the "Body Shrinking" filter— 22 despite long-standing evidence from Musical.ly that such filters distort self-image and promote harmful 23 body ideals, particularly among young, school-aged girls. Ex. 632A at 174:14–18, 181:8-18, 181:15-18; 24 Ex. 695. These filters remain available on TikTok's website today, accessible to users of all ages without 25 restriction or warning. Ex. 632A at 196:1-9; Ex. 623. 26 27 28 104 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 117 of 235 1 2 3 4 5 6 7 8 9 10 11 12 Figure 2. 13 14 15 16 17 18 19 20 21 22 23 24 Figure 3. 25 That same year, TikTok began testing default-on beauty effects—automatically retouching users' 26 faces, including those of minors, without notice or consent. The effects whitened teeth, smoothed skin, 27 and altered facial features such as eyes, lips, and nose. Ex. 633 at 112:5-113:21, 160:13-161:11, 170:8- 28 105 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 118 of 235 1 171:4; Ex. 624 at 2309-2310, 2315, 2319. These changes were implemented without consulting the Trust 2 & Safety team and in direct contradiction to expert guidance. Id.; Ex. 633 at 149:18-151:25. 3 Inside TikTok, the Trust & Safety team repeatedly warned leadership about the dangers of beauty 4 filters, but its concerns, as always, went unheeded. One member, , told 5 that influencing product teams was "challenging" because safety recommendations "risk[ed] negatively 6 impacting UG [User Growth] metrics." at 149:10–14, 152:4–153:20; Ex. 625 at 3295. Another 7 confirmed that new product changes did not routinely undergo safety review. Ex. 633 at 131:24-132:10. 8 By 2024, admitted that TikTok lagged industry 9 standards in addressing the harms of beauty filters to young users. at 121:4–122:6; Ex. 626 at 10 7555–56. Yet TikTok never warned parents, educators, or schools about the harms posed by its filters or 11 the safety risks it took in developing them. Ex. 444 at 494:25-498:8; Ex. 418 at 109:21-110:6. 12 (e) Algorithm 13 TikTok built its recommendation system for one purpose: to keep users—especially school-aged 14 children—on the app as long as possible. Ex. 531 at 212:18-213:13; Ex. 505 at 0457 (the "main goal" is 15 to maximize "user retention and stay duration"). 16 Ex. 427B at 467:1-468:6; Ex. 504 at 8330. 17 Ex. 427B at 18 505:20-506:7; Ex. 503 at slide 21; Ex. 502 at 9725. The formula is simple: more engagement means more 19 ads—and more profit. Ex. 427A at 134:2–13. 20 In addition to causing compulsive use, the engagement-driven model causes other harms. It traps 21 users in algorithmic "rabbit holes," repeatedly serving similar content based on content-neutral user 22 behaviors ( ). Ex. 23 507 at 7196; Ex. 427B at 519:23-520:10. TikTok's safety leadership knew these loops caused harm. In 24 early 2020, external experts warned , that rabbit holes fueled body 25 image issues and even a "contagion effect" around suicide. at 217:25-218:15; Ex. 509 at 4763 26 ("No doubt about it"); cf. Ex. 466 (results of "a massive (N = 689,003) experiment on Facebook" in 2014 27 showing "that emotional states can be transferred to others via emotional contagion"); Ex. 982 ¶¶ 123- 28 125. Although TikTok could identify users caught in these cycles, Ex. 507 at 7199, management 106 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 119 of 235 1 "deprioritized" mitigation efforts and stripped them of resources. Ex. 402A at 245:15-246:10; Ex. 510 at 2 0555. 3 After The Wall Street Journal's 2021 exposés, TikTok responded—not to protect users, but to 4 protect revenue. Ex. 633 at 261:12-262:19, 264:14-265:2, 267:25-269:8, 278:14-280:8; Ex. 511 at slide 3. 5 One investigation found that of videos served to a new "depression" account related to depression 6 content. Ex. 427A at 416:8-11, 416:25-417:24; Ex. 512 at 0484. Another reported "teen girls sucked into 7 eating disorder/dieting rabbit holes," where TikTok content "triggered" or "accelerated" their conditions. 8 Ex. 531 at 167:23-168:5, 168:17-169:13; Ex. 513 at 0824. TikTok's own data confirmed that of 9 eating disorder videos viewed by minors were algorithmically recommended. Ex. 533 at 250:7–20, 252:1- 10 21, 253:14-18; Ex. 515 at 1706. 11 Advertisers began threatening to pull , prompting TikTok to act— 12 not for user safety, but to protect revenue. Ex. 402A at 244:8-24; Ex. 510 at 0554 (discussing "[w]hy do 13 we need to solve this problem"). TikTok's , proposed adding "some 14 diversity" to users' feeds, "but not too much, otherwise users will leave (retention will drop)." at 15 417:25-419:24; Ex. 512 at 0486. TikTok's "filter bubble" fix was capped at a reduction in time 16 spent—any more was unacceptable. Ex. 427 at 403:10-25, 427:21–428:13; Ex. 514 at 7749. When eventually 17 implemented in 2022, the solution was only effective. Ex. 633 at 281:18-283:16; Ex. 516 at 9678. 18 TikTok never addressed suicide or self-harm rabbit holes—because The Wall Street Journal had 19 not reported on them. Ex. 633 at 270:21-271:5, 276:17-277:5, 278:14-279:4, 374:21-376:13; Ex. 511 at 20 slide 3; Ex. 517 at 8485 ("Ease of entering a filter bubble of [suicide/self-harm] content in U.S." remains 21 "HIGH"). Even after determining with "high confidence" that a user had taken his own life following 22 exposure to suicide content, TikTok did not act. Ex. 633 at 341:18-344:4; Ex. 518 at 5925. 23 Later, TikTok introduced a "reset" option to refresh feeds, but an internal document acknowledged 24 the flaw: "a [15-year-old] shouldn't have to" rescue themselves from an algorithmic rabbit hole. Ex. 512 25 at 0485. They also considered a "non-personalized feed" to reduce harm—but implemented it only in 26 Europe, because "ad revenues [would] decrease" if it was offered in the United States. Ex. 519 at 1580. 27 Despite its knowledge of these risks, TikTok never warned children, parents, or schools that 28 TikTok relentlessly optimized its algorithm for engagement and engagement alone, or that this single- 107 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 120 of 235 1 minded focus on engagement as expressed through users' behavior (time spent, swipes, likes, etc.) could 2 push minors into rabbit holes that put them at risk of eating disorders, depression, and self-harm. Ex. 633 3 at 386:17-387:18. 4 Despite its knowledge of these risks, TikTok never warned children, parents, or schools that 5 TikTok relentlessly optimized its algorithm for engagement and engagement alone, or that this single- 6 minded focus on engagement as expressed through users' behavior (time spent, swipes, likes, etc.) could 7 push minors into rabbit holes that put them at risk of eating disorders, depression, and self-harm. Ex. 633 8 at 386:17-387:18. 9 (f) Notifications 10 TikTok's notification system is intentionally designed to drive engagement, not to protect users. 11 Internal documents describe "constant notifications" as a "coercive design tactic" that "detract[s] from 12 user agency[.]" Ex. 541 at 9176. TikTok uses notifications to repeatedly draw users—particularly 13 students—back to the platform. Ex. 411 at 273:2-18 (push notifications appear even when the app is 14 closed); Ex. 411 at 272:18-273:1, 274:10-17 (notifications used "to reengage inactive users"); Ex. 528 at 15 9604 ("From the platform perspective, push notifications . . . Engage users[;] Drive activeness: directly 16 awake users[;] Lengthen stay duration: better engage users to stay in the app"). The objective was explicit: 17 "encourage users to open the App more and stay longer." Ex. 529 at 9759. 18 Notifications are not purely automated. Some are manually written and sent by employees of 19 TikTok, and their timing and volume are strategically optimized to 20 Ex. 529 at 9765; Ex. 411 21 at 274:2-9, 282:8-283:2. TikTok actively encourages users to enable notifications, Ex. 565 at 4506, and 22 even when users opt out, many continue receiving them. Ex. 646 at 2 ("I TURNED NOTIFICATIONS 23 ALL OFF! but I'm still getting them!"); Ex. 647 at 2 (user reporting notifications despite disabling them); 24 Ex. 648 at 2 ("I have my notifications completely off . . . and yet I still have notifications"). 25 TikTok is fully aware of the detrimental impact of these tactics on minors. Internal research 26 revealed that children lose sleep due to the barrage of nighttime notifications—as many as 145 in a single 27 evening. Ex. 404 at 83:10-23; Ex. 603 at 8036. Before 2021, TikTok sent notifications during school hours 28 and late into the night, sometimes as late as midnight. Ex. 540 at 1075. The company's own Trust & Safety 108 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 121 of 235 1 team warned that the "compulsion to react to notifications at night" results in poor sleep quality and 2 "impacts our minor users significantly." Ex. 602 at 1842. 3 TikTok considered limiting notifications to reduce disruptions to minors' sleep and academic 4 performance, including default restrictions during school hours. Ex. 413B at 207:4-9; Ex. 572; Ex. 643 at 5 8753. Internal analyses warned, however, that such limits would harm "business metrics," causing a 6 decline in daily active users. Ex. 643 at 8753; Ex. 572 at 8046. TikTok therefore declined 7 to disable notifications during school hours. Ex. 413B at 274:13-291:25; Ex. 643 at 8753. As one employee 8 admitted, "The only reason" the company did not limit school-hour notifications "is because of the 9 priority." Ex. 413B at 249:1–10; Ex. 573 at 5063. Nor did it implement default settings to restrict use 10 during the school day. Ex. 413B at 198:21-25, 199:1-6. 11 Despite knowing that these notifications disrupted students' sleep and concentration, TikTok never 12 warned schools, parents, or teachers about their frequency or impact on learning. Ex. 413B at 178:20- 13 179:4, 223:23-224:2, 224:8-13; Ex. 444A at 291:3-292:3. 14 In short, TikTok weaponized notifications to keep children constantly engaged—during class, late 15 at night, and at the expense of their well-being. Even after internal teams warned that these tactics 16 disrupted sleep and learning, TikTok refused to act, prioritizing engagement metrics over minors' safety. 17 By knowingly maintaining practices that foreseeably harmed students and schools, TikTok once again 18 chose profit over responsibility. 19 (g) TikTok Now 20 In 2022, TikTok launched TikTok Now—a feature prompting users to "snap and share front and 21 back camera shots in real time" within a three-minute window after receiving a notification. Ex. 638 at 22 1183. The company proceeded despite warnings from the Trust & Safety team, which raised "significant 23 concerns from a Minor Safety perspective." Id.; Ex. 637; Ex. 640 at 6143 ("[TikTok Now] will diminish 24 our ability to keep teens safe."). The team urged restricting the feature to users 18 and older, but leadership 25 rejected that recommendation, admitting that "[t]he whole point of launching [TikTok Now] is that it is 26 attractive to U18." Ex. 639 at 1944.00006; Ex. 641 at 3556 ("we need to consider business growth in such 27 cases even with lower safety"). 28 109 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 122 of 235 1 Ignoring Trust & Safety, TikTok rushed to release TikTok Now to compete with BeReal, described 2 internally as "the fastest growing product in the history of the internet." Ex. 634 at 4365. Predictably, after 3 launch, minor safety became "the biggest issue risk that surfaced from [TikTok Now]." Ex. 641 at 3556. 4 Even before launch, TikTok's Trust & Safety team warned that TikTok Now would disrupt 5 classrooms, as it required immediate user engagement following synchronized notifications—often during 6 the school day. Ex. 633 at 392:6-397:3; Ex. 642 at 8648 ("Teens will receive notifications to engage with 7 TikTok Now during school hours."). The team cautioned that "schools and teachers are going to hate" the 8 feature because "kids already have smartphone addiction in class," and adding real-time prompts "with a 9 synchronized push notification and a very real sense of FOMO" would be "a recipe for trouble." Ex. 405B 10 at 359:14-360:5; Ex. 634 at 4364. 11 Those concerns were quickly realized. TikTok's post-launch data revealed "significant posting" 12 from children ages 13–17 during school hours—triggered by the platform sending TikTok Now 13 notifications in the middle of the school day. Ex. 637 at 1557–58; Ex. 644 at 9718 ("Currently teens 14 receive TikTok Now notifications during the school day"); Ex. 418 at 231:2-17, 233:10-24, 235:9-14, 15 235:23–236:3 (confirming that responding to TikTok Now notifications in class disrupts instruction). 16 These outcomes mirrored pre-launch warnings that the feature would drive posting "during school hours" 17 and "during sleep hours." Ex. 637 at 1558. But at no point while this feature was being developed or after 18 it was launched did TikTok warn parents, educators, or schools, about the educational disruption TikTok 19 was creating. 20 (h) TikTok LIVE 21 TikTok launched its live-streaming feature, TikTok LIVE, around December 2019 as part of a 22 broader effort to expand engagement and monetization among younger users. Ex. 619 at 41-42. TikTok 23 LIVE gave school-aged users—as young as sixteen—the ability to broadcast their lives to a global 24 audience with almost no safeguards in place. Id. 25 Employees quickly saw what that meant. By early 2022, concerns about predation and exploitation 26 on TikTok LIVE had grown so serious that TikTok initiated an internal review—Project Meramec—to 27 examine how minors were using the platform and assess the risks it posed. Ex. 418 at 85:22-86:19, 148:6- 28 13. The findings were alarming: children and high school–aged teens were hosting livestreams and 110 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 123 of 235 1 receiving digital "gifts" from adult viewers—gifts that could be converted into real money. Ex. 677 at 65; 2 Ex. 678 at 8273; Ex. 679 at 6385-6386; Ex. 418 at 118:6-120:22, 122:22-123:8, 154:19-161:12, 201:6- 3 205:19. Many of these minors, including those under 13, had accessed TikTok LIVE by signing up through 4 third parties such as Google, Facebook, or Instagram, bypassing TikTok's age gate entirely. Id. TikTok's 5 own data confirmed that school-aged users were being exposed to grooming and exploitation—and that 6 the company was profiting from it. Id. 7 , who led Project Meramec, tried to stop it. She warned that TikTok LIVE posed 8 serious dangers to children and teens—ranging from exploitation to addiction—and escalated her concerns 9 directly to senior executives, including 10 . at 131:23–133:18; Ex. 680 at 1509. 11 documented that "senior leaders know about this issue" and that it was "right up front for 12 in late February." at 138:17-139:11, 149:23-150:2, 150:8-16; Ex. 685 at 8271. 13 Leadership, however, refused to act. As later explained, she had "been fighting about fixing this 14 for months with people who didn't think the harms are severe enough." at 151:11-19; Ex. 685 at 15 8271. 16 TikTok never publicly released the Project Meramec data, or warned of the risk of sexual abuse 17 that LIVE created. Ex. 444 at 494:25-498:8. Instead, parents, educators, and schools would have to wait 18 for the press. 19 In April 2022, Forbes published "How TikTok Live Became 'A Strip Club Filled With 15-Year- 20 Olds,'" exposing the same dangers had identified. at 79:21–82:17; Ex. 682. The article 21 detailed how underage users performed sexual or suggestive acts for adult viewers in exchange for 22 TikTok's virtual gifts. Id. at 2. Internally, confirmed that "[e]verything this Forbes article is 23 addressing are things we already knew" and admitted she had been "on a crusade to fix this problem for 24 months now." at 131:23-133:18, 135:4-15, 135:20-136:10; Ex. 680 at 1509. She later 25 acknowledged that the publicity may have been "a good thing" because it "put pressure on leadership to 26 finally take action to protect kids the way [she] was recommending." at 140:25-141:4. 27 and her colleague, , expressed relief that media 28 scrutiny might succeed where internal advocacy had failed. wrote, "The only silver lining in press 111 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 124 of 235 1 like this is that they do tend to spark a fire," while hoped the article would "put[] a fire behind other 2 potential solutions I've been advocating for." at 138:17-139:8, 140:16-142:4; Ex. 680 at 1509. 3 Yet their optimism was tempered by frustration. As another employee put it bluntly: "[P]roduct never 4 takes safety as a priority till things blow up. [N]o amount of complaining helps!" Ex. 418 at 180:24-182:7; 5 Ex. 686 at 1290. 6 (i) CSAM reporting 7 TikTok knew it had a serious problem with child sexual abuse material ("CSAM") on its platform. 8 Internal data showed it was "twice as likely to find child sexual exploitation on TikTok than [its] 9 competitors," not because there was more of it, but because TikTok was "facilitating access to CSAM." 10 Ex. 683B at 279:20-280:16, 281:10-282:20; Ex. 687 at 7494. At the same time, TikTok knew it was 11 "under-reporting CSAM" to the National Center for Missing and Exploited Children ("NCMEC"), despite 12 acknowledging internally that "TikTok has a responsibility to report CSAM" to that organization. Ex. 13 683A at 228:19-229:10; Ex. 683B at 282:23-283:15; Ex. 687 at 7494.00001. 14 This failure was not accidental—it was the predictable outcome of TikTok's flawed reporting 15 system. Reporting CSAM was so unintuitive that users were effectively discouraged from doing it. In 16 August 2019, employee told the company's , that to 17 report CSAM, a user first had to press the "share" button—an action typically used to send videos to 18 friends, not to report criminal content. at 49:18-50:5, 61:2-62:21; Ex. 689 at 2348-49; 19 at 231:21-232:10 (confirming "it's not intuitive for a user to press the share button when they come across 20 child pornography"). Even after pressing "share," the reporting option was hidden—"listed on the far right 21 of the sharing options," requiring users to scroll to find it. Ex. 521 at 49:18–50:5, 61:2-62:21; Ex. 689 at 22 2348-49. 23 The problem persisted into the following year. Reporting remained "buried under the share 24 button." Ex. 683A at 234:16-235:1. Members of TikTok's Trust & Safety team admitted that the process 25 was "horrible," acknowledging internally that "[i]f we want to be serious about safety and give our words 26 some credence, we need to allow people to easily report problems." Ex. 683A at 230:9-16, 231:2-232:6, 27 233:3-22; Ex. 690 at 4; Ex. 691 at 1102 ("Clear/intuitive reporting: TT sucks here"). TikTok's leadership 28 recognized the dangers and the failures in its own systems, yet did nothing—ensuring that users remained 112 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 125 of 235 1 unable to easily report serious risks or violations and left parents, educators, and schools in the dark as the 2 problem festered. Ex. 444 at 494:25-498:8. 3 (j) TikTok's Buried "Safety Center" 4 acknowledged that it was essential for 5 TikTok to provide users, parents, and the public with clear and easily accessible information about how 6 to stay safe on the platform. at 227:24-228:11. But when TikTok employees tried to find that 7 information themselves, they couldn't. The Safety Center—the page that was supposed to guide parents 8 and teens through TikTok's safety tools—was so deeply buried that even employees "thought we just 9 didn't have any of this information after ten minutes of searching around." Ex. 683A at 237:24-238:11 10 (agreeing it is "not reasonable" to expect people to spend that much time searching to "see if TikTok has 11 information on safety"). Internal documents confirmed what everyone already knew: "The additional steps 12 to link to our 'other' help center is awkward and really buries information"; the Safety Center was "not 13 sufficient to meet the needs of our users"; and "[w]e have an obligation to be a lot more on top of this and 14 own our end of the bargain." Ex. 683A at 235:15-237:13, 238:12-239:6; Ex. 684 at 6665-66. 15 The problem wasn't just neglect—it was a choice. Even TikTok's safety videos, created to educate 16 users about online risks, were "never shown," making them functionally invisible. Ex. 683A at 230:20– 17 231:1. TikTok's supposed safety resources were as hard to find as its willingness to prioritize them. 18 (k) Barriers to Exit 19 TikTok deliberately engineered a coercive system to addict young users—and just as deliberately 20 trapped them when they tried to leave. Having built a platform designed to capture attention and drive 21 compulsive use, TikTok ensured that even those seeking to escape its grip would face a maze of obstacles. 22 Every step of the user experience, including account deletion, was crafted to maximize retention and 23 minimize departures, which in turn increased profits. 24 TikTok's failure to provide effective tools to manage compulsive use has driven many users— 25 especially minors—to try deleting their accounts altogether. Ex. 672A at 159:18-20, 160:4-10, 162:2-13, 26 167:24-168:5, 169:2-7, 175:9-17; Ex. 583 at 2 (user reporting, "it got me addicted and now I have to delete 27 it"); Ex. 586 at 2 ("I've come to the decision to delete my account"); Ex. 587 at 2 ("Trying to get over my 28 TikTok addiction. Please delete."). But TikTok intentionally made that process difficult. Deleting an 113 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 126 of 235 1 account requires navigating through ten separate steps—a system deliberately engineered to discourage 2 users from leaving. Ex. 669 at 17 (images depicting the ten steps needed to delete an account). 3 In 2021, TikTok received roughly account deletion requests per day. Ex. 670 4 at 9388. Rather than viewing this as evidence of user harm, it treated it as a retention problem and added 5 new barriers "to make sure only users who really want to delete [an] account submit the deletion requests." 6 Id. TikTok implemented two deterrents: a "pre-check" reminder prompting users to complete tasks before 7 deletion, and a "questionnaire" asking why they wished to delete their accounts. Id. These tactics worked 8 as intended—by December 2021, of users quit the deletion process midway, most abandoning it at 9 the questionnaire ( ) or data download screen ( ). Ex. 671 at 9722. 10 From those questionnaires, TikTok learned that many users sought only temporary deletion—often 11 citing "temporal leave" ( ) or "other reason" ( )—and immediately began exploring ways to "save" 12 those users by offering temporary deletion options and re-engagement strategies. Id. Even users who 13 completed all ten steps could not easily leave. TikTok automatically placed deleted accounts into a 30- 14 day reactivation period, during which logging in even once canceled the deletion process. Ex. 672A at 15 311:20–312:2, 313:10–22; Ex. 645; Ex. 669 at 13. TikTok knew users struggling with addiction were 16 especially vulnerable during that time. Ex. 672A at 310:10–19, 311:2–15, 314:8–315:15, 322:6-19, 17 322:23–325:1, 326:25-327:16, 327:20–328:15; Ex. 673 at 2 ("I have already deleted the account several 18 times."); Ex. 674 at 2 ("Don't give me 30 days because I'll always come back to it."); Ex. 675 at 2 ("Please 19 delete my account immediately. I am unable to wait the 30 days because I am addicted to this app."). Yet 20 even when users explicitly requested immediate deletion, TikTok refused to shorten or waive the 21 reactivation period. Ex. 672A at 314:3-5. 22 (6) TikTok's misrepresentations and manipulations 23 TikTok's leadership actively buried the findings of the company concerning user harm. TikTok 24 maintained , while publicly issuing 25 statements claiming that its platform promoted "creativity, self-expression, and joy." Internal studies were 26 never released, and public-facing materials like the "TikTok Guide for Parents" made no mention of the 27 known and documented mental health, sleep, or educational harms. 28 TikTok's consistent response to evidence of harm was to manage perception, not risk. It 114 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 127 of 235 1 commissioned studies to identify problems, then silenced or reframed the results to protect engagement 2 metrics and corporate image. In doing so, it perpetuated the very harms its researchers—both internal and 3 external—warned of, proving that these consequences were not only foreseeable but foreseen. 4 (7) TikTok failed to warn parents, educators, and schools about 5 known mental health risks 6 As detailed above, TikTok knew that its platform placed children and schools at grave and 7 foreseeable risk. It knew that school-aged users were being harmed by compulsive use, sleep loss, body- 8 image distortion, eating disorders, depression, anxiety, and exposure to self-harm content. Yet it chose 9 silence—failing to warn the very parents, educators, and institutions responsible for protecting those 10 children. 11 TikTok failed to warn students, parents, educators, school administrators, and school districts that 12 use of TikTok may be harmful to the mental health of school-aged children. Ex. 413B at 173:3-14. Nor 13 did TikTok warn these groups that use of the platform increases the risk of serious harms, including 14 addiction, compulsive or excessive use, depression, anxiety, suicidality, self-harm, eating disorders, body 15 dysmorphia, and sleep disorders. Ex. 413B at 172:7-11, 172:16-173:1, 173:14-21, 173:23-174:4, 174:6– 16 12, 175:8-15, 175:17-25, 176:2-7; Ex. 444A at 293:18-296:1. 17 Similarly, TikTok failed to warn teachers, school administrators, and school districts that students' 18 use of TikTok foreseeably harms schools. Ex. 413B at 177:8-13; Ex. 444A at 289:18-290:1. TikTok did 19 not disclose that schools would likely need to divert staff time and resources to: (1) confiscate cell phones 20 used for TikTok during school hours; (2) respond to threats made against schools and students through 21 the TikTok platform; and (3) implement enhanced technology measures to restrict student access during 22 school hours. Ex. 444A at 296:24-297:8, 297:10-17, 298:3-11. Nor did it warn that student that TikTok 23 use during class time disrupts instruction and distracts both students and teachers. Ex. 413B at 177:15-20, 24 178:4-13; Ex. 444A at 290:2-9. 25 Moreover, TikTok failed to inform schools that students can post on TikTok and receive 26 notifications during school hours—and that such notifications interfere with study and concentration, an 27 issue TikTok's own teams had already recognized. See supra § (f) (Notifications); Ex. 413B at 178:20- 28 179:4, 223:23-224:2, 224:8-13; Ex. 444A at 291:3-292:3. Finally, TikTok failed to warn parents, 115 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 128 of 235 1 educators, and schools that, as a foreseeable result of student TikTok use, 2 schools would likely need to: (1) hire additional mental health staff; (2) train 3 staff about the harmful effects of social media; (3) develop educational 4 materials addressing social media addiction; (4) divert staff time to 5 confiscate cell phones during school hours; (5) repair property damage 6 linked to TikTok-related incidents; and (6) increase technology resources to 7 restrict access to TikTok. Ex. 444A at 292:5-293:16, 296:3-9, 296:25-297:8, 8 298:3-11. 9 Despite knowing the harms its platform inflicts on students and 10 schools, TikTok continued to market itself as an educational tool, boasting 11 that "millions of people learn something new on TikTok." Ex. 696. Its ads— 12 such as the one featuring "Mrs. Kelly" depicted to the right (Figure 4)— 13 underscore the deception, celebrating supposed educational value while concealing the platform's harm 14 to children, educators, and schools. 15 c) Snap's Conduct 16 (1) Snapchat's business model 17 Snapchat is a social media platform that catapulted to popularity with teens shortly after its launch 18 in 2011. Ex. 800 at 7142 (Snapchat users are "primarily a younger demographic"). Snap generates revenue 19 primarily by advertising to its users. Ex. 829 at 51:14-18 (advertising was Snap's principal method of 20 generating revenue since 2014). Snapchat advertisements appear across the app, in between posted Stories, 21 on the public Discover and Spotlight video feeds, within the camera, when browsing augmented reality 22 lenses, and through the Snap Map. See Ex. 921 at 1031 (slide depicting where Story Ads appear); Ex. 807 23 at 39:17-40:11, 81:20-82:14 (describing appearance of ads in Discover feed and on Spotlight); Ex. 1150 24 at 7128 (ads appear in Stories, camera, Discover, and in commercials); Ex. 829 at 54:3-14 (ads appear in 25 Snap Map). To increase revenue, Snap aims to increase daily active users, increase engagement, and to 26 increase the number of active advertisers. Ex. 921 at 1043. 27 Snap ads are specifically designed to increase daily user time. See 1151 at 0291 (design plan 28 explaining that "ads will be more closely linked together, making a better user experience for all and 116 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 129 of 235 1 keeping Snapchatters on the platform for a longer period of time"). Snap also created profiles of its users 2 so advertisers could target by age, gender, location, language, lifestyle, shopping preferences, and 3 numerous other characteristics. See Ex. 807 at 153:16-162:24. By 2018, Snap offered advertisers self- 4 service interfaces so they could seamlessly create ads using these characteristics without interfacing with 5 a Snap representative. See id. at 24:7-16. 6 Teens are core to Snapchat's business model. See Ex. 827 at 112:13-24 (head of product research: 7 "our core audience is teenagers."); Ex. 807 at 127:11-16 (former product team lead: "[Snapchat is] the 8 number one social media app for teenagers in the United States."). Snapchat dominates in terms of teen 9 market penetration—by 2014, Snap boasted that it "absolutely own[s] the youth demo in the US," with 10 roughly 84% of the market among users aged 10 to 24. Ex. 825 at 4207. Ahead of Snap's IPO three years 11 later, investors observed that middle school children under the age of thirteen were "rabid . . . user[s] of 12 Snapchat." Ex. 826 at 7064; see Ex. 866 at 7658-59 (email sent from Snap's head of communications to 13 CEO reporting that 40 percent of children under 13 use Snapchat). By 2023, this number had grown, with 14 Snapchat reaching "90% of the 13- to 24-year-old population[.]," Ex. 822 at 4550. That same year, CEO 15 Evan Spiegel estimated that about 90 million users were teens. Ex. 803 at 140:21-141:7. (As discussed 16 below, this number was likely far higher given that many teens report their age inaccurately.) 17 Further, users aged 13-17 represent nearly 50 percent of the time spent on Snapchat. Ex. 833. 18 Internal documents characterized Snap as a "daily habit" for almost 50 percent of "extremely engaged" 19 13-to-17-year-olds in the U.S. Ex. 823 at 1034 ("Young users are extremely engaged and send Snaps 20 daily"); see Ex. 831 at 6185 ("this generation can't live without our app"). And, as Snap's vice president 21 of Product testified, users spending more time on the app generates more revenue. See, e.g., Ex. 829 at 22 63:17-64:17; Ex. 828 at 7675 (Snap "revenue formula" turns on "time spent" per monthly active user). 23 Snap has powerful incentives to capture users as young as possible. Internal analyses showed that 24 the 13-to-17 age cohort generated the highest average revenue per user (ARPU) during the first 12 months 25 following registration. Ex. 828 at 7745-746; Ex. 829 at 75:24-77:6. The economic advantage of early 26 acquisition is clear: a senior product designer testified that acquiring a user at age 13 would "produce 27 revenue for Snap at age 13, 14, 15, 16, 17, and 18, and then forward," whereas users joining later in life 28 generated much less revenue. Ex. 829 at 80:5-81:2 (agreeing that "it's to Snap's benefit to onboard users 117 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 130 of 235 1 as early as possible to maximize the amount of time that a user is in this 13 to 17 demographic where they 2 are . . . the highest revenue-producing demographic on Snapchat"); see Ex. 830A at 215:18-21 (Senior 3 Director of Product Management: teenage users who "start using Snapchat when they're 16 or 17, a really 4 high proportion of them continue using that service"). Internal presentations explicitly emphasized the 5 importance of targeting the youngest allowable users, labeling the cohort "very valuable for advertisers" 6 and emphasizing that the "Snapchat Generation has $2 trillion in purchasing power, making up over 40 7 percent of global consumers." Ex. 831 at 6185; id. at 6182 ("continuing to win with new 13 year olds is 8 the most critical aspect to onboarding new users"); Ex. 829 at 104:11-105:5; Ex. 991 ¶¶ 154-156. Snap 9 also claimed that, through teens, advertisers could reach their parents, because the "Ads [teens] saw on 10 Snapchat influence what they asked for and what their parents purchased." Ex. 902 at 8460. 11 Snapchat has seen tremendous success. Revenue increased by an average of 21.2% per year from 12 2020 through 2024, hitting $5.4 billion in 2024. See Ex. 836, Ex. 837, Ex. 838, Ex. 839, Ex. 840, Ex. 841 13 (10-K and 10-Q Reports). Importantly, a significant percentage of Snap's profitability is attributable to 14 teenage and adolescent users; from 2020-2023, Snap earned over $1 billion in revenue each year from the 15 under-18 age cohort. Ex. 991 ¶¶ 164-65. 16 (2) Snapchat's targeting of school-aged children 17 Not surprisingly given its business model, Snapchat is designed to appeal to teens and adolescents. 18 Ex. 817 at 3471 (internal research explaining that "childish perception of Snapchat derives from the 19 stereotype that that [sic] most Snapchat users are young people"). The platform appears in bright colors, 20 primarily yellows. Its logo is a cartoonish ghost. Ex. 818 at 9163 (adult users observe that "the overall 21 aesthetic of the app feels cartoonish"). Further, users are represented on the platform via "Bitmoji"— 22 avatars with childlike proportions that allow each user to present themselves as a cartoon version of anyone 23 they choose. Ex. 819 at 5958 ("Bitmoji Stories have become the Sunday cartoons"). Figure 5 is a 24 representative depiction of Snapchat's bitmoji-driven aesthetic, from its internal documents (Ex. 815 at 25 0094). Snap uses Bitmojis to infer user gender for "content and ad ranking, suggestions, targeting, [and] 26 optimization." Ex. 820 at 5893 (describing plan to use Bitmoji data to improve gender inference for 27 advertising and noting "[g]ender is one of our strongest user profile attributes"). During the school year, 28 118 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 131 of 235 1 Snap's goal is to "[k]eep Bitmoji top-of-mind among teen-skewing lifestyle media and introduce them to 2 the many ways they can use Bitmoji to personalize their back-to-school style." Ex. 821 at 2488. 3 4 5 6 7 8 9 10 11 12 13 14 Figure 5 (from Ex. 815 at 0094). 15 Snapchat also makes use of numerous other features that appeal primarily to teenage sensibilities, 16 including ephemeral messaging, gamified features like Streaks, and augmented reality lenses (some of 17 which are playful and cartoonish). Upon launch, Snapchat's most distinctive characteristic was that video 18 or photo messages, or "Snaps," were ephemeral, and vanished shortly after they were received. Ex. 801 19 at 661:5-663:10; Ex. 802 at 0241; see also Ex. 803 at 89:18-90:22; Ex. 804 at 5803 (Snapchat was created 20 as an app whose primary purpose was to allow users to send each other photos that would automatically 21 disappear). Snap touted its disappearing messages as a way for "young people" to "combat" the pressure 22 to "curat[e] a perfect image of themselves on social media." Ex. 816 at 2006. But the dark side of this 23 feature was made clear almost from its inception—Snapchat "quickly gained a reputation as a sexting 24 platform." Ex. 805 at 4149. 25 While other platforms copied Snapchat's signature features like ephemeral messaging and Bitmoji, 26 Snapchat itself evolved by adopting key characteristics common to other major social media platforms. 27 See Ex. 834 at 96:25-97:17 (referencing numerous features common to other social media platforms). As 28 119 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 132 of 235 1 with Instagram and TikTok, users are encouraged to enhance their appearance in video or photos by 2 applying beautification filters to smooth or whiten their skin, remove blemishes, and make them appear 3 slimmer. See Ex. 835 at 0990 (describing lenses to include: "Skin smoothing/whitening," "Acne/Blemish 4 removal," "Dark circling / Eye enlarger / Eye brightening," and "Slimming / Stretching"). At the same 5 time, many of Snapchat's lenses, like the platform's overall aesthetic, are overtly cartoonish, for example, 6 allowing "young people to put on bunny ears or [] vomit a rainbow." Ex. 801 at 466:10-467:1; Ex. 865 at 7 6224 ("Snapchat lenses are believed to be more childish than TikTok lenses overall"). Users are also 8 presented with an algorithmic video feed that mimics TikTok's endless scroll. See Ex. 808 at 3050 ("We 9 already have an endless scroll design in Discover and I think we wish it was more addictive compelling.") 10 (strikethrough in original); Ex. 807 at 311:21-312:1 (acknowledging similarities between TikTok and 11 Snapchat's Spotlight); see also Ex. 834 at 96:24-97:17. And the videos they post can receive "likes" or 12 "favorites" that signify the social approval of their peers. See Ex. 807 at 297:20-22. 13 Snap has innovated on the idea of social communication by transforming it into a gamified 14 experience—encouraging users to send daily Snaps to friends, acquaintances, and strangers in their friend 15 list simply to maintain a "Snap Streak" with those users, see Ex. 801 at 193:22-194:3; share with other 16 users where they are—and what they are missing—at all times via a map feature, see Ex. 809 at 8084 17 ("problem statement deck" identifying "FOMO – seeing a party happening on Snap Map" as a "negative 18 social interaction"); and broadcast content to friends by posting video Stories, see Ex. 807 at 37:17-40:24, 19 which, like Snaps, are ephemeral and thus prompt users to constantly check the platform before Stories 20 disappear. See Ex. 1152 at 2064 ("The ephemeral nature of Stories requires users to develop a daily habit 21 of checking them"). Snap provides each user with a "Snap Score" that awards more points for greater 22 engagement with the app. Ex. 810 at 3228 (describing Snap Score as "feeding into the gamification 23 mindset among Snapchatters"); Ex. 811 at 4720 (Score functions to "reward and incentivize users" for 24 virtually every platform interaction—sending Snaps, adding friends, posting Stories, or maintaining 25 Streaks). Snap Trophies rewarded users for achieving Snap Score milestones, engaging in large volumes 26 of chats, utilizing certain filters, or viewing prescribed amounts of video content on Snapchat's endless 27 video feeds, Discover and Spotlight. Ex. 814 at 4019-20, 4053-60. And users are incentivized to add an 28 ever-expanding cohort of "friends," most of whom they do not know in real life. See Ex. 806 at 6130 120 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 133 of 235 1 (teens use group "Shout Outs" to increase "the visibility of their profile" so that "loads of people" will 2 "add them" as friends). Snap added this last feature to encourage users to further increase their Snap Score, 3 enmeshing each user in deeper engagement with the platform. 4 Snap understands that Snapchat's interconnected reward systems create addictive usage patterns 5 disproportionately affecting adolescents. Ex. 810 at 3227 (acknowledging that Snapchat "is social 6 gamification, which tends to be popular amongst the younger demographics."). Snap expressly recognized 7 the psychological impact of the Snap Score, for example, acknowledging that "[e]ducating users" about 8 their Snap Score creates an "addictive" effect that "encourage[s] them to produce and consume more 9 snaps," transforming them into what Snap calls "high-quality users." Ex. 812 at 8652. Snap also 10 understood that the gamified Score presented safety risks, incentivizing young people to "keep their 11 privacy settings open, accepting strangers' [friend requests], and participate in shout outs and streaks in 12 order to gain more contacts and views on Snapchat and eventually increase their score." Ex. 983 ¶ 93. 13 Given its numerous design choices to target teenagers, Snapchat struggles to attract and retain 14 users who are in their mid-twenties and older. See, e.g., Ex. 823 at 1034 ("In US, Snap sending is an 15 everyday . . . behavior for young users – but not so for older users"); Ex. 830A at 190:1-191:3 ("Gen X 16 and older users generally have never used Snapchat"). Thus, in order to maintain growth, Snap must 17 constantly drive users to Snapchat while they are young to offset the loss of users who age out of the 18 platform's teenage-focused ecosystem. Senior Snap executives have acknowledged the importance of 19 continually recruiting new teenage users— with Spiegel explicitly asking, "Should we acquire new 13 20 year old users?" and CFO Derek Anderson emphasizing that continuing to "win with new 13yr olds" is 21 "super key" and "the most critical aspect to onboarding new users." Ex. 831 at 6182-83. Conversely, when 22 confronted with data showing "a decline in the US 13-to-17 trend line," Boyle wrote that he felt 23 "absolutely sick to my stomach," while Spiegel and other executives expressed "great concern" at reports 24 of "a dramatic drop in 13 to 17" year-old users. Ex. 842; Ex. 843 at 7523; Ex. 834 at 28:4-21. 25 (3) Snapchat's targeting of schools 26 The financial imperative to relentlessly recruit young users led Snap to deliberately target and 27 infiltrate schools. Snap understood that schools provided a crucial avenue for growth. High school students 28 121 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 134 of 235 1 in particular were characterized by Snap as its "core target audience" and "instrumental" to growth. Ex. 2 855 at 5097, 99; see also Ex. 856 at 1476. During back-to-school periods, Snap observed large spikes in 3 new friendship, finding that users add "47% more friends from Group Chats during back to school 4 months." Ex. 857 at 0369. This surge early in the school year provided "a tailwind for engagement for the 5 remainder of the year," and led to sustained use throughout the school year. id. at 0359-60. 6 Accordingly, Snap leaned especially hard into the back-to-school period, consciously seeking to 7 "make Snapchat synonymous with back-to-school" and, specifically, the "purchasing behaviors" 8 associated with that time of year. Ex. 867 at 97:15-98:2; see also Ex. 830A at 246:5-247:6 (acknowledging 9 that Snap considers back-to-school season to be a "really important time of year" and that it is a 10 "commonly understood period in time where there's additional marketing"). Snap created "back to 11 school" messaging to convey to teenagers and adolescents that Snapchat "is the app for students." Ex. 860 12 at 0064. Snap claimed to be "part of the family during back-to-school." Ex. 902 at 8443. Snap also 13 strategized on how to "[g]et High Schoolers attention" by leveraging its features, such as filters and lenses, 14 to create a "high intent audience," as "over 90% of high schoolers are in the lens carousel on a monthly 15 basis." Id. at 8472, 74, 79. The lens carousel included school-related offerings such as a "School Vibes" 16 and "School Style" filter. Ex. 904; Ex. 905. 17 Meanwhile, Snap told advertisers that teenagers used Snapchat for every action leading up to their 18 back-to-school purchases: 21% looked for inspiration in Stories, 33% used the app to get advice from 19 friends, nearly 75% were interested in "trying" new products using Snapchat's augmented reality features 20 before buying, and 27% took Snaps using in-store filters. Ex. 902 at 8468. Snap created numerous 21 additional avenues for advertisers to reach teenagers at school, including full-screen Snap Ads and 22 commercials on the Discover tab, each with the goal of creating "instant impact" and to "start building a 23 qualified audience" of teenagers at school. Ex. 903 at 0337. And, as described in more detail below, by 24 2016, Snap obtained precise high school location data to enable advertisers to geotarget students on 25 campus, boasting to advertisers that Snapchat "rules the halls of high school." Ex. 807 at 134:24-135:7, 26 138:19-140: 24; Ex. 903 at 0334; see also Ex. 903 at 0032 ("Snapchat reaches 2x as many 13-17 year- 27 olds in the US than Facebook, Instagram, and Messenger – combined"). 28 122 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 135 of 235 1 Snap's intentional targeting of schools is woven into its history. Shortly after Snapchat's 2011 2 launch, Evan Spiegel wrote in an official company blog post that he was "thrilled to hear that most of 3 [Snapchat's users] were high school students who were using Snapchat as a new way to pass notes in 4 class—behind-the-back photos of teachers and funny faces were sent back and forth throughout the day." 5 Ex. 824 at 2; Ex. 803 at 111:14-113:14. Spiegel added that the app's usage skyrocketed during school 6 hours and otherwise waned. Ex. 824; Ex. 803 at 111:14-113:14. 7 In 2017, Snap's Chief Strategy Officer provided Spiegel with an investor analysis titled, "Snap: A 8 Middle School Crush," which described "the middle school set [as] a rabid – almost exclusively – user of 9 Snapchat." Ex. 826 at 7064. Internal company documents from 2018 showed that 64% of Snapchat users 10 aged 13-21 reported using Snapchat during school. Ex. 845 at 3471. This was consistent with outside 11 research performed on Snap's behalf, which showed that Snapchat accounted for 39-40% of the total 12 battery usage for high school students, while it accounted for 31% for middle school students. Ex. 846 at 13 5372-73; see also Ex. 1000 ¶ 39 (explaining that 45% of phone use at school "was for Snapchat alone"). 14 Snap also surveyed teenage users and knew that teens "use Snapchat to communicate with friends" at 15 school "unless the teacher tells them not to." Ex. 849 at 7235; Ex. 827 at 470:10-472:2. Snap personnel 16 referred to students' Snapchat use during school as "under-the-desk" phone time, and Snap witnesses 17 confirmed they were aware that students used Snapchat on their phones during class and took pictures of 18 events at school. Ex. 851 at 416:10-417:24; Ex. 852 at 6439; Ex. 853 at 367:14-368:1. In short, Snap has 19 long known that Snapchat is frequently used in classrooms during school hours across the United States, 20 and it encouraged that use. Ex. 854 at 1969. 21 To further boost student engagement, Snap created a feature called School Communities to connect 22 students at the same school by class year. Ex. 857 at 0372. Each School Community was a closed space 23 where students could share school-related Stories and send messages to one another. Ex. 861 at 124:19- 24 125:20. The point of the School Community feature was to "grow[] the user base" by targeting high school 25 students and providing them a place to congregate. Ex. 864 at 8680. At a 2023 company town hall, Snap 26 personnel boasted that its School Community feature had "15M students across 24K US high schools!" 27 Ex. 857 at 0371; 861 at 125:1-20. Because of its growth potential, School Community caught the eye of 28 Snap's competitors, with Meta personnel internally remarking, "Our competitors in social connection are 123 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 136 of 235 1 leveraging school integrations as a way to keep in touch with their community." Ex. 258 at 7145; see also 2 Ex. 269 at 5745 (Instagram presentation focusing on School Communities and observing, "Our 3 competitors are exploring school communities"). 4 Following rollout of this growth-centric feature, Snap "learned that some of the largest public 5 school districts ban external emails, preventing students from receiving the identity verification emails" 6 from Snapchat, which were necessary for students to join their School Community group. Ex. 864 at 8679. 7 Rather than respect these external email bans, Snap created alternative verification methods to circumvent 8 the email restriction. Id. Likewise, Snap developed technology to circumvent other proactive efforts by 9 schools to ban or limit Snapchat use in school. During COVID, when students were participating in web- 10 based learning, senior Snap personnel noted "under the desk" phone time was decreasing because children 11 weren't in school, and recommended that the company turn its attention to developing a web-based 12 product. Ex. 851 at 417:3-419:11. At Spiegel's direction, Snap developed Snapchat for Web, allowing 13 students to access the platform from a laptop or tablet to avoid discipline for using smartphones during 14 class. Ex. 934 at 6632 (explaining that Spiegel directed the development of a "fun and engaging laptop 15 experience" so that kids could use Snapchat "in class"). In addition, Snap prepared for schools to 16 eventually ban cellphone use altogether, designing a wearable device on which students could use 17 Snapchat during class should they be prohibited from using their phones. Ex. 952 at 4134-35. These 18 deliberate workarounds show that Snap not only knew its platform disrupted schools but actively 19 engineered ways to ensure that disruption continued. It is especially ironic that Defendants now cite 20 COVID as an alternative cause of school disruption when Snap's own records show it treated the pandemic 21 as an opportunity to deepen student engagement and maintain classroom use. 22 (4) Snap's knowledge and willful blindness of harms 23 Although Snap's relentless pursuit of teen engagement has been good for the company's growth, 24 Snap knows that use of Snapchat can be harmful for minors. And while the relentless penetration of 25 Snapchat into schools has likewise helped Snap's bottom line, Snap understands that use of Snapchat at 26 school is disruptive, interferes with the educational environment, and harms schools. Snap has long known 27 enough that it should have warned users and their caretakers, including parents and schools. Instead of 28 doing so, Snap went out of its way to maintain a stance of plausible deniability. 124 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 137 of 235 1 Snap's co-founders, Spiegel and Bobby Murphy, have been aware since as early as 2013 that users 2 experience what they termed "Snapchat addiction." Ex. 922 at 4154 ("frankly the ones that have the 3 Snapchat addiction have no room for anything else. Snaps dominates their life."). Over the years, young 4 users regularly reported to Snap that they were distressed from their compulsive use of Snapchat, with 5 some reporting they felt addicted and admitting to opening the app several hundred times per day. E.g., 6 Ex. 928. In 2018, employees casually discussed articles describing Snapchat's "elements of addictive 7 design [that] encourage high user engagement." Ex. 923. These articles were consistent not only with user 8 reports but with Snap's internal findings that Snapchat had become a "daily habit" for nearly 50% of 9 "extremely engaged" 13- to 17-year-old daily active users in the United States. Ex. 924 at 1034. 10 Snap understands that addictive and habitual use is driven in part by specific Snapchat design 11 choices. For example, because the Snapchat platform is built around ephemerality, it is designed to "drive 12 a phenomenon commonly called FOMO—Fear of Missing Out. Knowing that a funny video or a party 13 photo will vanish by tomorrow compels teens to open the app frequently, so they don't 'miss' what their 14 friends are doing. It creates a 24/7 urgency to be online." Ex. 989 ¶ 84; Ex. 927 at 2045 ("The ephemeral 15 nature of Stories requires users to develop a daily habit of checking them."). 16 Meanwhile, Snapstreaks and the Snap Score gamify the Snapchat user experience in a way Snap 17 understood to be addictive and anxiety-provoking. See Ex. 812 at 8652 ("Educating users what score is 18 and how to increase score may make users care about (if not become addicted to) their score and thus 19 encourage them to produce and consume more snaps, and ultimately become high-quality users"). The 20 Snapstreak feature, launched in 2015, requires users to exchange Snaps daily with specific friends to 21 maintain numerical counters displayed with fire emojis; if a user fails to send a Snap on a given day, the 22 Streak is lost. Snap recognized the feature's compulsive nature, characterizing this mechanism as "Daily 23 Habit Training." Ex. 925 at 6161. Further, it acknowledged that the Snapstreak feature "tapped into mass 24 psychosis" where "people must keep the streaks going." Ex. 935 at 9344; see also Ex. 987 ¶ 143 25 (describing Streaks as Snap's "'playing by appointment,' compelling users—especially teens—to return 26 to the platform daily in order to avoid losing [the Streak]"). Internally, Snap's product team expressed the 27 "stance on streaks…that we don't love them (it was an accidentally addictive, somewhat unhealthy feature 28 that gamifies friendship in a weird way)." Ex. 926 at 9271. "[B]ut they're too delicate to touch right 125 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 138 of 235 1 now"—given that over fifty million users had Streaks. Id. In 2018, Spiegel himself asked senior product 2 managers for a plan "to kill streaks" because "it's the right thing to do." Ex. 971 at 6015. But Streaks was 3 never killed, because of its impact on teen engagement. Ex. 1153 at 1495-96 (referring to streaks as 4 "compulsive behavior" that is "a main driver of . . . engagement"). 5 Streaks play an outsized role in causing younger users to experience anxiety and stress. Ex. 933 at 6 5093 (internal engineering document describing Streaks as a "self-perpetuating, anxiety driven system" 7 with "compulsive behavior pattern[s]"); Ex. 989 ¶ 69 (same). This is because Streaks "take advantage of 8 adolescent susceptibility to social cues by providing a key indicator that (a) an individual adolescent has 9 social connections and (b) that they are in contact with their social connections daily….This is vitally 10 important to adolescents in terms of their social development." Ex. 983 ¶ 87. Nonetheless, Snap designed 11 Streaks to target its core teenage demographic, with senior personnel acknowledging it is a feature 12 primarily used by teen users. Ex. 847 at 307:6-308:21. The feature's penetration among high school 13 students demonstrates its effectiveness: 41% of high school Snapchat users maintain an average of 9.5 14 Streaks simultaneously, demanding constant attention throughout the school day to avoid losing their 15 accumulated counts. Ex. 936 at 6711, 15. Losing a Streak can cause teenagers and adolescents enough 16 anxiety that many contact Snapchat customer support to plead for reinstatement of lost Streaks. Ex. 929 17 (email to Spiegel from high school student stating, "I can't imagine how many emails you get per day 18 from teenagers threatening you if you don't return a streak they lost"). As explained by the Districts' 19 expert, Dr. Tim Estes, "Snap could have, and should have, turned this feature off for minors. Instead, Snap 20 doubled down, and monetized it." Ex. 989 ¶ 72. Specifically, in 2023, Snap monetized teen stress by 21 offering a paid option to restore a user's Streak. Ex. 937 at 0645-47 (early launch internal data predicting 22 a "floor" of in annual revenue from Streak restore). 23 Finally, Snap knew that its platform's "addictiveness" affected student performance at school. Ex. 24 930 at 5370. It learned this is true not only because of compulsive use during the school day, but because 25 of the disruptive impact Snapchat has on young users' sleep, which affects student performance. Ex. 938 26 (internal distribution of research demonstrating negative impact of compulsive social media use by teens 27 on sleep). After learning this, Snap conducted no further research into whether Snapchat was affecting 28 student's academic performance and simply buried its head in the sand. Ex. 847 at 225:18-227:13. 126 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 139 of 235 1 Indeed, the company ignored repeated requests over years to assist schools in handling the unique 2 disruptions caused by Snapchat use at school. Teachers have repeatedly contacted Snap requesting that it 3 block the use of Snapchat during school hours or remove geofences around their school because of 4 concerns related to student harassment or sharing of CSAM. See, e.g., Ex. 979 at 7542; Ex. 853 at 376:4- 5 379:2; Ex. 980 at 9005; Ex. 813 at 91:24-92:9, 99:17-100:21; Ex. 873 at 5839-5840; Ex. 801 at 689:22- 6 692:25. Likewise, school administrators have complained directly to Snap's leadership that they are 7 "slammed with Snapchat issues," and the app is a constant "distraction." Ex. 920 at 4453; Ex. 801 at 8 708:17-709:20. Snap knew this was true—its research observed that children use Snapchat at school to 9 "cure boredom," Ex. 911 at 3999; Ex. 861 at 111:18-112:24, and it knows that, when bored students are 10 distracted by Snapchat, that would "disrupt the teacher's ability to teach." Ex. 851 at 37:11-38:1. But Snap 11 has done nothing in response to these complaints and requests. 12 At the 2023 National Student Safety and Security Conference, Snap's Head of Global Platform 13 Safety, Ms. Beauchere, met with educators and school resource officers to discuss schools' experiences 14 with Snapchat and how the app was being used in school environments. At the panel, "[a]ll participants 15 said that Snapchat is part of their everyday life at school and said both parents and students regularly 16 report to them harms or concerns related to Snapchat." Ex. 801 at 696:1-700:1. These educators and school 17 resource officers described concerning interactions facilitated by Snapchat, including its ephemeral 18 messaging, which encouraged teens to engage in risky behavior like sharing CSAM. See Ex. 920 at 4451 19 (describing parents and students informing schools about "harms or concerns connected to Snapchat" and 20 connecting "perceived ephemerality" to, among other things, "sharing of nudes and financial sextortion 21 of males"). One educator called Snapchat "the #1 cause of drama in school aged children." Ex. 920 at 22 4452; Ex. 801 at 706:7-14. Snap heard these concerns and, again, did nothing. 23 Given its extensive external information documenting potential harms to teenage users, as well as 24 information it learned from user experience surveys, Snap could have and should have changed its harmful 25 product features and, at the very least, warned users, parents, and schools. As detailed in the subsections 26 below, it did neither of these things. 27 In addition, Snap made a deliberate corporate decision not to conduct meaningful internal research 28 into whether Snapchat was harming the mental health of its adolescent users. Morgan Hammerstrom, who 127 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 140 of 235 1 leads Snap's sole user experience and product research team, testified that she was never asked to research 2 users' experiences related to mental health. Ex. 827 at 96:22-25. Further, although Snap conducted user 3 surveys to gauge how Snapchat lenses made all users (not just teenage users) feel, no one at Snap ever 4 requested research into whether users experience body image issues from using Snapchat or whether the 5 platform increases user anxiety or stress. Id. at 97:4-13. When directly asked whether it mattered to her as 6 the head of Snap's user experience research if users were experiencing body image issues or anxiety 7 related to Snapchat features, Hammerstrom testified that it did not. Id. at 97:14-98:5. 8 This research void is not incidental but systematic. Hammerstrom could not identify a single 9 instance where mental health had been included as a question on any survey or interview that her team 10 conducted with Snapchat users. Id. at 166:4-12. Out of thousands of research projects conducted or 11 overseen during her tenure at Snap, including a period when she was the company's only research 12 employee, Hammerstrom has never researched whether users find Snapchat addictive, the platform's 13 impact on youth mental health, anxiety, or depression, or the impact of features like endless scroll and 14 ephemerality (not to mention increased time spent) on teens. Id. at 606:3-608:7. 15 Most revealing, Hammerstrom testified that investigating user harm falls entirely outside of her 16 job responsibilities. This is notable, as Hammerstrom led Snap's product research department and was 17 involved in all research the company conducted. When asked whether she would want to know if users 18 were addicted to Snapchat, she testified that such information "wouldn't have anything to do with [her] 19 job." Id. at 151:3-9. She further stated that if a user reported Snapchat was addictive, her team would not 20 investigate that concern but would instead focus exclusively on product-specific feedback. See id. at 21 105:7-19 (describing an example of her research as assessing reactions to whether "Snapchat buttons were 22 purple"). To Snap's head of user research, teenage harm was simply not her concern. 23 Other senior Snap personnel confirmed that the company systematically ignored or avoided 24 conducting any research on the impact of its platform on teen or adolescent (or anyone's) mental health. 25 When asked whether Snap collects data relating to mental health impacts, David Boyle, Snap's Senior 26 Director of Product, deferred to Hammerstrom. Ex. 830A at 174:13-175:11. He could identify no instance 27 in which he requested any study to be conducted or data collected related to Snapchat's impact on teen 28 mental health. Id. at 175:13-18. Ms. Beauchere, Snap's global head of platform safety, acknowledged that 128 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 141 of 235 1 Snap never performed research on whether any feature of the app is addictive, despite internal company 2 conversation suggesting that both Spotlight and Streaks encouraged compulsive use. Ex. 801 at 451:9- 3 452:5; see Ex. 973 at 3224-25, 44-45 (internal document identifying infinite scroll and autoplay as 4 "unhealthy gaming mechanics" and observing that users "feel obligated" to maintain Streaks, which 5 "become[s] stressful"). 6 Juliet Shen, Snap's former Lead Safety Product Manager, testified that mental health was 7 "volunteer work" at Snap because the company did not have a dedicated product manager focused on 8 mental health and well-being. Ex. 948 at 105:11-18. Ms. Shen confirmed that she could not recall any 9 research on whether using lenses or filters, posting stories, checking or monitoring Stories and Spotlight 10 views, using Snap Map, maintaining and monitoring Streaks, monitoring Snapscore, using trophies or 11 charms, or being encouraged to add friends on Quick Add impacted teen mental health. Ex. 948 at 189:6- 12 202:12; Ex. 962. In discovery, Snap admitted it has no unit or department specifically dedicated to 13 researching whether its features cause teen users to suffer depression, body dysmorphia, or compulsive 14 use of Snapchat. Ex. 963 at 22-25. Spiegel, Snap's CEO, also admitted under oath that he had never 15 commissioned research to determine whether or not, or to what degree, Snapchat was endangering youth, 16 and acknowledged that Snap never employed in-house child psychologists, neuroscientists, or mental 17 health professionals to advise the company on matters related to the mental health of young people. Ex. 18 803 at 47:8-14, 48:1-14, 50:5-13, 108:3-12. 19 Snap knew enough about the harmful effect of its features, their addictive potential on teenagers 20 and adolescents, and the disruption its platform was causing in schools to merit a warning. Yet, again, 21 Snap not only failed to do that, it attempted to remain willfully blind of the problems it was causing. 22 Numerous high-ranking Snap personnel testified they never conducted, ordered, or recalled any analysis 23 or in-depth research into Snapchat's impact on the academic performance of high school students or 24 teenagers. See, e.g., Ex. 847 at 226:20-227:13; Ex. 829 at 122:16-24. Nor was research conducted with 25 teachers, school employees, or counselors to better understand Snapchat's impact on schools. Ex. 827 at 26 100:5-12; Ex. 948 at 330:18-23. In short, "Snap's approach is characterized by what has been described 27 in behavioral ethics as the 'ostrich syndrome,' where one chooses to bury their head in the sand and ignore 28 moral responsibilities and the related standards of care. The ostrich syndrome creates what scholars have 129 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 142 of 235 1 labeled as 'moral myopia,' in which one's moral responsibilities do not come clearly into focus, and it can 2 even result in moral blindness." Ex. 993 ¶ 390. 3 (5) Snap's failure to exercise reasonable care 4 The Snapchat platform, as designed, functions as an ecosystem that encourages compulsive and 5 harmful use by its adolescent users. Features like Snap Score, Streaks, and ephemerality, described above, 6 gamified communication and compelled teens to obsessively monitor Snapchat to bolster their in-app 7 persona and ensure they did not miss out on Stories, messages, and videos posted by their friends. 8 Snapchat's Quick Add feature prompted users to recklessly accept friend requests from strangers to 9 increase their Score, connecting unsuspecting teens with dangerous child predators. Snapchat's camera 10 incorporated beauty lenses that drove body dysmorphia, disordered eating, and anxiety about appearance. 11 And Snapchat adopted the addictive innovations pioneered by its social media peers, including endless 12 algorithmic feeds, predatory notifications, and geotargeting. The sum effect of these features was to drive 13 addiction and mental health harm among adolescents—and that was by design. 14 Because Snapchat relies so heavily on adolescents to drive and maintain growth, Snap has never 15 implemented effective age verification. Snapchat's practically non-existent age gate allowed millions of 16 users to misreport their age and access the platform before they were thirteen years old. Worse, until 17 2022—the year this litigation was commenced—Snapchat provided parents with no parental controls at 18 all. Even now, those "controls" are highly ineffective, not promoted by Snap, and as a result, scarcely 19 used. As a result, Snapchat's youngest users, including millions who are not supposed to be on the platform 20 at all, are exposed to a bevy of harmful features. 21 (a) Age Verification 22 Snapchat's process for verifying a user's age at sign-up has always been, and is, completely 23 ineffective. Ex. 871 at 8461; Ex. 867 at 285:5-286:5 (senior director of product observing that Snapchat's 24 registration flow resulted in "scenarios where we're highly confident the inputted age is incorrect."). Prior 25 to 2013, Snapchat did not require users to enter their age at all during registration. Ex. 867 at 143:16- 26 144:7. By 2015, Snap still did not know the age of approximately 25 million daily active users who were 27 not required to submit their age at signup. Ex. 868 at 4875. When Snapchat finally introduced a birthday 28 field in its registration flow, that field defaulted to the year 2000, a setting that automatically made users, 130 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 143 of 235 1 including those under 13, appear to be teenagers regardless of their actual age. Ex. 869B at 785:4-787:5; 2 see also Ex. 1149 at 5673 (internal document stating that Snap personnel were unsure when this change 3 was implemented). Snap modified its registration process in 2017 to default a new user's age to 18, which 4 was no improvement since any child who failed to input their birthday would appear to be 18. Ex. 870; 5 Ex. 863 at 118:11-14 (confirming that "if a user doesn't manually enter their real birthday they are 6 automatically set to 18"). And while Snap fixed this issue in other countries, Snapchat in the United States 7 to this day defaults a user's age to 18. See Ex. 863 at 173:17-174:1, 184:11-185:16. 8 At no point in its existence has Snapchat required users to do anything meaningful to verify their 9 age when creating an account. So long as new users input a birthdate identifying themselves as 13 or 10 older—or inputs no birthdate at all, thus defaulting to 18—they will successfully pass Snapchat's age gate. 11 See Ex. 863 at 393:24-395:14 (demonstrating that Snapchat's account creation flow allows users to create 12 an account with a fake name, birthday, and email). This frictionless age gate effectively encourages young 13 users to misrepresent their age at sign-up—including users under the age of 13, who Snapchat's terms of 14 service nominally prohibit from using the platform. Ex. 987 ¶ 137. 15 Senior Snap personnel, including its Global Head of Platform Safety Jacqueline Beauchere, 16 acknowledge that children—especially those under 13—will misreport their age at sign-up to gain access 17 to the platform. Ex. 801 at 482:7-484:1; Ex. 874 at 0931. Worse, when children under 13 do accurately 18 report their age—and are denied registration—Snapchat permits them to immediately re-register on the 19 same device using a different birth date. Ex. 863 at 216:16-223:15; Ex. 867 at 162:5-164:14 See also Ex. 20 874 at 0931 ("when a user tries to register and fails because they input an age under 13," they are "allowed 21 to register again" on the "same physical device"). Snap's product teams considered prohibiting this 22 practice and requiring "locked out" periods but ultimately opted against it. Ex. 869B at 855:6-856:13. 23 According to Senior Vice President of Global Policy, Jennifer Stout, lockout periods "would have a 24 negative impact on daily new users and daily active users," causing a "material impact to user acquisition 25 and growth." Ex. 878 at 1089. 26 Snap does not take proactive measures to find underage accounts or remove them when they are 27 detected. Snap has never conducted an audit to determine the number of users under thirteen that maintain 28 Snapchat accounts. Ex. 863 at 95:7-17. The company does not track or otherwise monitor whether ages 131 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 144 of 235 1 reported on its platform are accurate. Id. at 251:9-16; Ex. 867 at 117:6-9. Nor has Snap conducted user 2 research into the number of children who were under the age of 13 when they created their accounts. Ex. 3 876; see Ex. 867 at.117:6-9. Snap's public Transparency Reports do not disclose the number of under- 4 thirteen accounts the company moderates because the numbers would be "laughable." Ex. 877 at 3323 5 (internal communication between product managers stating that, whereas TikTok deletes 20 million 6 underage accounts per quarter, Snap "delete[s] like 50 per day" and obscures this data in its Transparency 7 Reports). In December 2021 alone, Snapchat had over 43 million users identified as under 13. Ex. 875. 8 During this same period, Snap deleted only 2,430 users—0.00005% of those identified as underage—for 9 accessing the platform before they were 13. Ex. 874 and 0931. 10 Not surprisingly, Snap knows that underage users are on Snapchat during elementary and middle 11 school, providing Snap an even earlier vector for growth. Ex. 851 at 421:7-16; Ex. 885 at 3460. Indeed, 12 Snap's Head of User Research, Morgan Hammerstrom, acknowledged Snap was aware of children as 13 young as the first grade using Snapchat at school. Ex. 827 at 451:5-452:20. One elementary school teacher 14 reported that students had been creating "anonymous" Snapchat accounts, "add[ing] a bunch of friends, 15 and then spread[ing] negative or not-appropriate content around the school," which a Snap employee 16 described as "not unlike 'note-passing' at scale without accountability." Ex. 885 at 3460; Ex. 851 at 421:7- 17 16. Reports such as this were shared with dozens of Snap personnel. Ex. 885. At a National Student Safety 18 and Security Conference, school resource officers met with Ms. Beauchere, Snap's Global Head of 19 Platform Safety, to explain that children as young as 9 and 10 years old were sharing CSAM over Snapchat 20 during their third and fourth grade classes. Ex. 920 at 4453; Ex. 801 at 712:11-713:3. Yet still, Snap did 21 next to nothing to proactively remove or moderate these underage accounts. 22 Notably, Snap has developed the capability to accurately estimate a user's age after registration— 23 but it has only used these technologies for advertising purposes, not to detect and remove users under 24 thirteen. See Ex. 807 at 161:16-162:23 (the purpose of Snap's age inference technology was "to target 25 ads"). Snap has developed several methods to estimate a user's age based on photos and videos they post, 26 as well as their interactions with other users and the platform. As early as 2015, Snap developed the 27 capability to estimate user age through facial analysis. See Exs. 868; 879. In 2016, Snap's age inference 28 132 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 145 of 235 1 technology was utilized by its monetization team to display products with age restrictions, like alcohol, to 2 appropriate age cohorts. Ex. 867 at 277:2-278:10, 279:24-280:18; Ex. 807 at 161:16-162:23. 3 Snap's machine learning model for its Spotlight platform can also infer when a user is under 13. 4 In response to fines imposed on TikTok in the UK, Snap "conducted an experiment to assess the volume 5 of potential under 13 escalations via our Spotlight Moderation workflow and escalated between 2,000- 6 3,000 Snaps per day globally," demonstrating it has the capacity to identify and moderate under-thirteen 7 accounts active on Spotlight. Ex. 884 at 4413. But again, that model is not used to flag underage users, 8 only to filter content displayed to certain accounts. Ex. 881 at 0935-36 (describing the purpose of and 9 ability to infer age below 13); Ex. 882 at 9464; Ex. 883 at 120:21-123:22 (describing the purpose of 10 Spotlight's age inference model to filter content, not flag underage users); Ex. 883 at 98:21-99:16, 120:5- 11 122:15 (former Head of Product explaining that "filtering out" from Spotlight means to "remove" or 12 "make not eligible" for viewing). In other words, although Spotlight's machine learning model can 13 identify underage users, Snap takes no action against those accounts. 14 Alex Osborne, Snap's former Safety Product Manager, attributed the failure to moderate underage 15 accounts on Spotlight to "operational bandwidth limitations on the [Trust and Safety] side." Ex. 884 at 16 4413. But the "bandwidth limitation" was deliberate. In 2023, Osborne proposed straightforward 17 solutions, including "simply increasing the staffing on our Trust & Safety team dedicated to age 18 verification" or "integrating with a third-party age assurance vendor that could verify user age through 19 selfies or ID." Id. at 4414. Snap rejected these proposals. Osborne explained the real concern: "there are 20 risks with this approach. For example, regulators might use our small experiment against us and demand 21 that we expand our usage of age estimation tools to other, more consequential parts of Snapchat." Id. As 22 of November 2024—over two years after this litigation commenced—Snap had not engaged with a third- 23 party age assurance vendor to moderate underage accounts on Spotlight. Ex. 863 at 114:21-115:16. 24 Snap could have utilized any one of the age estimation technologies it developed to identify users 25 under 13 and remove them from the platform, but it has failed to do so. Id. 863 at 258:11-260:17. Prior to 26 2024, the company's sole approach to identifying and moderating such accounts was to rely on reports or 27 complaints from other users. See Ex. 867 at 88:14-89:1. Even this reporting avenue was deliberately 28 weakened, however, because Snap does not permit users to report suspected under-thirteen accounts 133 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 146 of 235 1 through the app itself; instead all reports must be made through Snap's website, adding needless friction 2 to the reporting process. See id. at 249:23-261:4. Snap's seeming indifference to the presence of children 3 under thirteen on Snapchat is particularly concerning given that Snap's founder and CEO admitted under 4 oath that "children under the age of 13 are not ready to communicate on Snapchat." Ex. 803 at 168:6-9. 5 The company looks the other way because children are so critical to Snapchat's growth and profitability. 6 (b) Parental Controls 7 Snap has failed to provide parents with effective controls to prevent their children from using 8 Snapchat compulsively and/or at school. Despite Snapchat's extensive reach among teenagers and 9 adolescents, Snap did not introduce any features resembling parental controls until 2022, eleven years 10 after the platform was created. Ex. 801 at 259:17-260:7. At the time, Snapchat was "the only major 11 platform that does not yet have parental controls." Ex. 853 at 409:23-421:21. Even after these tools, called 12 "Family Center," were introduced in 2022, they were designed to be weak and serve more as a window- 13 dressing marketing tool as opposed to an actual parental control. See Ex. 889 at 9334 (Family Center 14 intended to be a marketing tool to "address parental concerns and school bans"). Ms. Beauchere instructed 15 her colleagues that she was "wary" to refer to Family Center as parental controls "because they're not;" 16 "traditional parental controls cover the who, what, where and how much of a young person's life," but 17 Family Center was not designed to provide parents those resources. Ex. 890 at 9241. 18 Family Center allows a child to link their Snapchat account to a parent's account, after which the 19 parent is able to observe certain actions taken by the child's account. Ex. 894 at 7950. As first introduced, 20 Family Center permitted parents to view only the names of their child's friends and the group chats they 21 were participating in, Ex. 1171 at 9123, but not to set time limits, restrict content, or control their child's 22 behavior on the platform in any meaningful way. Ex. 891 ("Time limits and other forms of restrictive 23 mediation are traditional 'parental controls.' Snap is not offering traditional parental controls[.]"); Ex. 893 24 at 9261 (Spiegel commenting, "I think the goal here is to give the parent visibility but not necessarily 25 control"); see also Ex. 890 at 9241 (Beauchere stating, "we've gone out of our way NOT to refer to them 26 as parental controls because they're not. Traditional parental controls cover the who, what, where, and 27 how much of a young person's online 'life.' These are built differently."). Although certain Snap personnel 28 considered allowing parents to control their child's privacy settings, Spiegel shot that idea down. See Ex. 134 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 147 of 235 1 893 at 9261, 9264; see also Ex. 801 at 281:15-282:-15. Snap has also declined to promote Family Center 2 to parents or teens, see Ex. 896 at 0984 (Snap had no plans for "community outreach"), and "Family 3 Center is extremely hard to find in the app," Ex. 894 at 7951. Snapchat does not prompt teenagers either 4 at sign-up or at any point to enable Family Center protections or inform teens that Family Center exists. 5 See Ex. 987 ¶ 139. 6 Because it is both hard to find and effectively useless, Family Center is not used by Snapchat users 7 or their parents. See, e.g., Ex. 894 at 7951-52; Ex. 801 at 303:22-304:12 (Beauchere acknowledging that 8 the number of users utilizing Family Center is "low"). Indeed, only 0.33% of the platform's teenage users 9 have utilized the feature. Ex. 894 at 7952. But adoption or promotion of Family Center was never the 10 point. Instead, the primary purpose of Family Center "was to have a feature we can point regulators, 11 interested parents, and the press to, rather than broadly tell everyone to adopt this." Ex. 897 at 6211. Why? 12 Snap was concerned that parental controls, including even weak pseudo-controls like Family Center, may 13 negatively impact "engagement . . . if this lands the wrong way with teens." Id. And Snap's overriding 14 goal has been to maintain its popularity with teens and adolescents at all costs. 15 (c) Barriers to Exit 16 When teenage and adolescent users decide to deactivate their accounts, Snap erects barriers. See 17 Ex. 898 at 2746 (explaining that Android users are unable to delete their accounts from in-app settings 18 and need to access the web portal instead); id. at 2756 (iOS users not able to delete account in-app until 19 January 2022); Ex. 899 at 4403 (inability to delete accounts on Android continued to persist into 2024). 20 Moreover, parents face significant hurdles when trying to delete the accounts of their underage children. 21 See, e.g., Ex. 900 at 9434 (communication from parent of underage child complaining that "Your website 22 doesn't allow parents to delete accounts"). 23 Once a user navigates the deactivation process and is about to delete their account, they are 24 presented with a black screen depicting a crying emoji and a message that reads, "Your account will be 25 deactivated, which means friends won't be able to contact you on Snapchat. You'll also lose any Chats 26 you've saved and Snaps and Chats you haven't opened." Ex. 901. The obvious goal of this message is to 27 discourage deletion. Finally, even a user who proceeds does not actually have their account deleted—Snap 28 135 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 148 of 235 1 "require[es] a mandatory 30 day 'cooling off' period during which the user could disrupt and cancel the 2 process at any time—simply by logging into the account again." Ex. 987 ¶ 151. 3 (d) Geolocation 4 Snap developed and deployed comprehensive geolocation infrastructure specifically designed to 5 identify, track, analyze, and grow use of Snapchat at schools across the United States. The foundation of 6 Snap's efforts was a proprietary geolocation database called "Verrazano," or "Snap's Places Database." 7 Ex. 906 at 6440; Ex. 907. Verrazano contained approximately 16,000 places categorized as "high schools" 8 in the U.S., with additional schools potentially categorized under "School" or "Education." Ex. 906 at 9 6456. Snap personnel described the purpose of this effort as creating "Snapchat world for high schools." 10 Id. To enhance this effort, Snap purchased school boundary polygons from third-party vendors, which the 11 company manually cleaned and supplemented. Ex. 907; Ex. 908. Snap also acquired granular location 12 data, including a comprehensive list of high schools' location data, from a vendor. Ex. 807 at 138:19- 13 140:24. This data allowed Snap's advertisers to geotarget students with a "sponsored spaces" geofilter 14 whenever those students were within a certain radius of the school. Id.; see also id. at 137:19-138:1 15 (explaining how geofilter ads generally work). In 2016, Snap charged advertisers $100,000 to access only 16 "17 high schools, which it systematically leveraged to increase ad dollars through school-by-school 17 penetration." Ex. 910 at 0698. 18 Snap invested in this geolocation infrastructure so it could understand students' every action while 19 in school and monetize that information. The company created a dashboard "to view the number of unique 20 users, geofilter swipes, and session counts within a geofence." Ex. 907 at 3631. This dashboard enabled 21 Snap to track real-time metrics for individual schools. Id. (demonstrating the tool's capabilities at an 22 employee's former high school and showing 1,400-1,500 unique users within the high school's geofence 23 during the school year). The analytics went beyond simple user counts. Snap calculated exact penetration 24 rates by comparing its user data against enrolled student populations. See id. at 3633 (reporting that "we 25 have 1,450 unique users out of 1,684 students for a penetration rate of 86.1%" at one high school). Using 26 this technology, Snap was able to "measure swipe rates and users within a fence to find areas in need of 27 geofilter," and could "easily see the student population of the high school and links to a detailed 28 breakdown of demographics, tests, and other information about the school." Id. 136 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 149 of 235 1 This school-focused approach had buy-in at the executive level. In September 2018, Josh Siegel, 2 then heading Snap Map and one of the company's top Product managers, articulated the company's growth 3 philosophy in an email to the growth team, proposing "better ways to model growth, specifically about 4 how we can model school-by-school penetration." Ex. 907 at 3633. Siegel's communications make clear 5 the strategic importance Snap placed on educational institutions: "US universities and high schools are 6 really our core, both in terms of historic growth and penetration, as well as the loudest source of user 7 feedback for us and the primary place investors/advertisers assume we are winning. If we lose these users, 8 we lose." Ex. 1169 at 3579 Siegel pushed for systematic monitoring of school penetration, requesting "a 9 regular report or analysis that looks at US school penetration" and confirming that Snap possessed "all the 10 data to figure out who these users are based on their GPS location or even age + IP-based location at the 11 minimum." Id. This school-by-school growth and surveillance strategy was fundamental to Snap's 12 expansion model. Siegel explained that understanding "school-by-school behavior and penetration" was 13 critical for determining "how to ignite growth in new markets by focusing on school-by-school takeover 14 tactics." Id. 15 There is no evidence in the record—anywhere—that Snap ever warned schools, administrators, 16 and teachers that the company was collecting detailed geolocation data about their students, for the 17 purpose of increasing their use of Snapchat during the school day. Snap's policies governing geofiltered 18 content targeting schools contain detailed guidance regarding material likely to be disruptive to the 19 learning environment, but nowhere do these procedures instruct employees to inform a school or its staff 20 that their location is geofenced. See Ex. 848 at 5121-30, 85, 97, 55. Further, although Snap has the 21 technology to geolocate and/or geofence schools, there is no evidence that Snap has ever used either of 22 these capabilities to prevent access to Snapchat by students that were on school property during school 23 hours. Ex. 829 at 328:23-329:17; Ex. 801 at 695:15-24. Indeed, Snap did just the opposite, ignoring 24 requests from school employees who realized their schools were geofenced and requested that those 25 geofences be eliminated. See, e.g., Ex. 1147 at 6480 (accusing Snap of "falsely identif[ying] our 26 institution's location to include surrounding neighborhoods not associated with our school or brand" and 27 "asking one more time for [Snap] to take this more seriously"); Ex. 1148 at 9022 (explaining a geofenced 28 Story "cannot be disabled"). 137 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 150 of 235 1 (e) Notifications 2 The ephemeral nature of messaging on Snapchat creates urgency to quickly view, respond, and 3 save communications, requiring users to develop daily checking habits. See Ex. 1154 at 2007 (ephemeral 4 messages cause "users to develop a daily habit of checking them"). Snap recognized that this "FOMO" 5 could be further manipulated with notification campaigns. Ex. 912 at 6474; Ex. 913. "Snapchat employs 6 a notification recapture strategy, leveraging social prompts to draw users back onto the platform through 7 prompts to view a friend's story, a new Snap, or a general prompt to catch up on content." Ex. 987 ¶ 140. 8 Indeed, Snap deploys push notifications as "a powerful lever for growth and engagement," sending 9 approximately 311 million notifications daily to drive reengagement. Ex. 914. In 2022, 10 notifications drove approximately app opens daily, with incremental daily active 11 users attributed to growth notification pushes. Id. 12 Snapchat's push notifications are one major source of distraction and disruption in the school 13 environment. Ex. 807 at 226:22-227:12. Snap specifically created notification campaigns targeting high 14 school students and school-based engagement, including "an aggressive back to school growth notification 15 plan that will include dynamic user targeting based on school start dates." Ex. 915. Snap's High School 16 Communities program, which reached 15 million students across 24,000 US high schools, deployed 17 targeted in-app notifications designed to drive student acquisition and activation, including messages such 18 as "Don't miss out – your school's shared story is taking off." Ex. 916 at 2516-17. Snap scheduled High 19 School Community notification prompts to be delivered to students at 8:00 a.m. and 2:00 p.m., to 20 encourage students to share content to their High School Community, and internally acknowledged that 21 this constituted a "dark patterns/nudging technique[,]" which garnered student attention and disrupted 22 schools. Ex. 917 at 7602-03. 23 It was not until 2023—one year after this litigation commenced—that Snap's Product research 24 team proposed "giving users the power to turn off notifications during school hours or when they should 25 be working or studying." Ex. 849 at 7231; Ex. 827 at 476:4-477:21. But even after Snapchat permitted 26 users to turn off notifications, the platform continued to prompt users to enable notifications. Ex. 830A at 27 256:20-258:2. Snap witnesses were unaware of any efforts made by the company, or even policies or 28 policy considerations, to restrict push notifications that adolescent users receive when they are likely to 138 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 151 of 235 1 be in school. See, e.g., Ex. 853 at 379:6-19. Snap's corporate designee testified that the company declined 2 to implement recommendations to disable notifications for student users during typical school hours. Ex. 3 861 at 283:12-284:19. Further, Snap did not "specifically allow" users to "schedule quiet periods such as 4 no notifications during school hours." Ex. 830B at 508:14-20; see also Ex. 1145 at 4826 (finding in late 5 2023 that the "main reason users disabled" all "notifications is because they want to focus on school, 6 work, or friends when they hang out with them" but Snapchat provides no option to quiet notifications 7 during specific time periods). Snap has presented no evidence on summary judgment that it ever warned 8 Districts, parents, or the public about the disruptive nature of its notifications. 9 (f) Screen Time Management 10 Snap has never implemented a time management feature or user control to limit notifications 11 during specific hours of the day; even when Snapchat provided users the option to control notifications, it 12 was either on or off. Ex. 830B at 502:11-504:19; see also id. at 508:14-20 (pushing responsibility for 13 disabling notifications onto operating systems). At one point, Snap implemented a feature allowing "quiet 14 hours" from midnight to 7 am, during which Snapchat would not send push notifications, but scrapped the 15 feature after it was characterized as an impediment to growth. Ex. 939 at 2437 (A/B testing tracks " 16 DAU increase" when feature removed). Although Snap considered "giving users the power to turn off 17 notifications during school hours or when they should be working or studying," Ex. 849 at 7231; Ex. 827 18 at 476:4-16, it did not do so. Snap still does not "specifically allow" users to "schedule quiet periods such 19 as no notifications during school hours." Ex. 830B at 508:14-20. Indeed, Snap's corporate designee 20 testified that the company declined to implement recommendations to disable notifications for student 21 users during typical school hours. Ex. 861 at 283:12-284:19. Further, senior Snap personnel testified that 22 the company does not restrict the ability of teenagers to use Snapchat during the seven hours they are 23 likely to be at school—and there is no evidence that Snap has ever taken any action, or implemented any 24 product design features, that would impede or prohibit the use of Snapchat on school property. Ex. 829 at 25 326:18-327:4; Ex. 801 at 695:2-14. Nor has Snap even investigated restricting the ability of a teenager to 26 access Snapchat during school hours. Ex. 853 at 370:20-371:11; see also Ex. 807 at 262:17-20 27 (acknowledging that Snap has never paused or blocked access to Spotlight during school hours). 28 Unlike certain of its co-defendants, Snap never implemented features that would encourage 139 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 152 of 235 1 teenage users to "take a break" or "timeout," or to temporarily pause their accounts. See Ex. 863 at 314:7- 2 318:7 (acknowledging that Facebook, Instagram, and TikTok allow users to pause their accounts but 3 Snapchat does not). Nor did Snap ever provide teenagers a warning that they were spending too much 4 time on Snapchat or that they should put the app down to learn or sleep. See Ex. 829 at 102:23-103:12; 5 see also Ex. 853 at 135:19-136:11 (acknowledging that Snap should provide warnings to educate its users 6 and enable them to better control their experience).Snapchat users requested the ability to pause their 7 accounts and Snap considered implementing it, Ex. 1146 at 0928, but ultimately never did. See Ex. 863 at 8 321:4-10. 9 (g) Algorithm and Infinite Scroll 10 Snapchat contains three primary algorithmic content surfaces, each engineered to maximize user 11 engagement through sophisticated ranking systems that predict and exploit user behavior patterns. These 12 surfaces—Stories, Discover, and Spotlight—function as distinct yet interconnected engagement 13 mechanisms that collectively trap users in a web of compulsive content consumption. 14 Stories is Snapchat's ephemeral surface displaying short videos posted by a user's friends. The 15 surface utilizes machine learning algorithms that rank friend content based on predicted engagement 16 metrics rather than in chronological order. Ex. 1155 at 7691. The algorithmic ranking for Stories relies on 17 user features including "age, gender, locale," and behavioral signals such as view patterns, interaction 18 history, and predictive engagement modeling. Id. at 7693-95; Ex. 859 at 221:4-222:22. The Stories 19 algorithm is paired with an auto-advance experience, which encourages users to spend as much time as 20 possible on the platform absorbing content. Ex. 950 (describing Stories redesign to create "[a]n infinite 21 feed of ranked, interesting stories," and "[d]rive an endless scrolling experience of interesting content"); 22 Ex. 829 at 107:6-111:18 (describing Stories as "an experience where you can scroll endlessly"). Driving 23 engagement on Stories is critical to Snap's business: "Increasing user engagement with the Stories tab is 24 critical for our business," as of Snap's ad revenue is generated from this tab." Ex. 1155 at 7960. 25 Discover is Snapchat's platform for professionally curated short videos featuring products or 26 experiences promoted by advertisers or media companies. Like Stories, the surface is an endless scroll of 27 algorithmically selected videos. See Ex. 808 ("We already have an endless scroll design in Discover and 28 I think we wish it was more addictive compelling") (strikethrough in original). The primary purpose of 140 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 153 of 235 1 Discover, like Stories, is to increase the time a user spends in the app. See Ex. 1156 at 0712 (internal 2 document identifying Discover as the "silver bullet" for engagement and observing that "Habits are built 3 in times of boredom. Instragram LIVES in this domain"). Snap personnel likened Discover to "feed-based 4 environment[s]" with "large and nearly 'endless' content supply," comparing the surface to "FB newsfeed, 5 Instagram, Twitter, and TikTok"—models that "work particularly well." Ex. 954 at 9605. Indeed, Snap's 6 engineers, when designing Discover, asked "what will most effectively create a daily habit? What best 7 takes users down engagement rabbit holes?" Ex. 1157 at 3871. 8 Spotlight, implemented in 2020, is Snap's attempt to replicate TikTok's scrolling video feed. Ex. 9 807 at 311:21-312:1 (acknowledging that Spotlight is similar to TikTok); Ex. 1158 at 8592 (users 10 immediately recognized Spotlight as similar to TikTok and Instagram Reels); Ex. 1159 at 4071 ("Snap 11 Inc. admitted that Spotlight was 'inspired by other platforms,' including TikTok"). Videos on Spotlight 12 are ranked using an algorithmic sampling model, paired with an endless scrolling and auto-advance 13 experience. Ex. 951 at 6698 (ranking team responsible for "reducing barriers to consumption."). This 14 algorithmic video feed dramatically increased engagement, with views increasing 65% and view time 15 increasing more than 99%. Ex. 985 ¶ 142. 16 Across all three surfaces, Snapchat's algorithms optimize for behavioral proxy metrics that drive 17 compulsive use rather than user satisfaction. See Ex. 1160 at 7572 (algorithm optimizes for "metrics like 18 view time, shares, favorites, subscribes, and replies"). Each algorithmic surface also utilizes an endless 19 scroll. Snap's senior personnel have not only characterized the endless scroll as addictive but admitted 20 that this was Snap's goal. Ex. 953 at 3050 (senior Snap personnel observing that they wished the "endless 21 scroll design in Discover" was more "addictive"); Ex. 954 at 9064 (internal goals for content feed: "Drive 22 increased depth of engagement, time spent and binging behavior"). Internally, Snap personnel worried 23 that their algorithmic feeds would further exacerbate what they acknowledged was a social media 24 addiction crisis with teens. See Ex. 955 at 7920 (developing communications strategy to respond to 25 potential question asking, "At a time when there are huge concerns about screen time—won't this fuel 26 addiction with your young users?"). Further, despite claiming to "optimize the viewer experience beyond 27 in-session engagement metrics," Snap admits it is "comfortable using in-session actions to proxy for 28 longer-term effects," revealing that its algorithms are, in reality, entirely short-term and engagement- 141 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 154 of 235 1 based. Ex. 985 ¶ 147. And the effectiveness of this algorithmic manipulation has been quantified in 2 Snapchat's own metrics showing that "Spotlight ranking [which is engagement-based] improved views 3 and view time +65% and +99% respectively, relative to a heuristic ranking holdout." Id. ¶ 142. Snap 4 further acknowledges it has "limited mechanisms for asking viewers how they feel" about their experience 5 on the platform and lacks "an understanding of long-term viewer and creator retention, and therefore 6 do[es] not optimize for it today." Id. ¶ 148. 7 Snap witnesses acknowledged that it should warn minor users if part of its platform posed a risk 8 to their safety. Ex. 250 at 133:1-14. They also acknowledged that if the company was aware of a safety 9 issue, it was obligated to educate the community and give them the tools to help users control their 10 experience. Id. at 135:19-136:11. They even agreed that part of these efforts "can absolutely consist of a 11 warning." Id. Yet, at no point did it warn users about the harms caused by its algorithms or infinite scroll 12 on Snapchat. 13 (h) Account Recommendations 14 Snapchat also uses algorithms to make friend recommendations through its Quick Add feature, 15 which launched in 2015 and was designed to maximize friending activity. Ex. 1161 at 9877 ("Quick Add 16 contributes to of friendship requests sent by new users, and is the primary lever for improving new 17 user retention"); Ex. 1162 at 4435 ("As of May 2022, an average of daily friend requests are sent 18 via Quick Add accounting for of daily friend requests"). The tool worked exactly as intended, 19 prompting teenagers to accumulate as many friends as possible to boost their Snap Score and Story views, 20 and to maintain Streaks. Ex. 1165 at 3854 (describing "unhealthy streakers" as maintaining "streaks with 21 a lot of random people . . . from Quick add"). By 2021, Snap personnel confirmed teenagers "actively 22 [choose] to keep their privacy settings open, accepting strangers' quick adds and participating in shout 23 outs and streaks in order to gain more contacts and 'views' on Snapchat and eventually increase their 24 'score.'" Ex. 806 at 6130. Internal documents recognized that teens' "Main motivation to having a lot of 25 friends is to increase Story Views and Snapscore," meaning that "[u]sers would add a bunch of people 26 from Quick-add." Ex. 1164 at 1318. 27 Unfortunately, Quick Add became a vehicle for child predation. Law enforcement agencies 28 globally reported that over 80% of child sexual grooming cases were associated with Snapchat. Ex. 1166 142 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 155 of 235 1 at 8468. When advocacy groups created test accounts as 13-year-olds, they "immediately got 2 recommendations for contacts in quick add who were strangers." Ex. 801 at 358:23-359:1. User reports 3 from October 2020 flagged concerns about "suggesting minors as friends in 'Quick Add.'" Ex. 830B at 4 616:10-617:1. As early as 2018, Snap knew that "grooming on Snapchat" occurred because of "gamifying 5 snap score" to add friends through Quick Add. Ex. 1167 at 6144987. Yet six years after launch, in 6 December 2021, Quick Add's algorithm still did "not utilize any age-related feature," freely 7 recommending adults to minors and vice versa. Ex. 956 at 7701. 8 In 2022, Snap purportedly attempted to alter the Quick Add algorithm to make it safer but those 9 changes proved largely ineffective. A 2023 study by Common Sense Media surveyed girls aged 11 to 15 10 and found that 57% had been contacted by strangers on Snapchat in ways that made them feel 11 uncomfortable. Ex. 1144 at 1228. When presented with this study, Snap's director of strategy testified that 12 this issue came up "again and again" in safety conversations regarding Quick Add. Ex. 853 at 300:12- 13 301:23. The fundamental problem is that Quick Add cannot function safely when Snap lacks reliable 14 information about users' actual ages. See Ex. 1170 at 8196-98 (Quick Add protections "rel[y] on knowing 15 the age of the user reliably" and citing an instance in which a child was sexually groomed through Quick 16 Add because she reported her age as 22). And, as set forth above, Snap has taken insufficient steps to 17 ensure it has reliable information about its users' actual ages. 18 (i) Beauty Filters 19 Snapchat implemented lenses in 2015, which allow users to smooth and whiten their skin, enlarge 20 their eyes, and incorporate other beauty enhancements on photos taken by a user. Ex. 940 at 0990-91. 21 Beginning in 2018, if not earlier, Snap possessed extensive evidence that its beautification filters 22 contributed to body dysmorphia and eating disorders among young users. See, e.g., Ex. 941 (Journal of 23 American Medical Association ("JAMA") article discussing the impact of edited selfies and photo filters 24 on perceptions of beauty and body image, which it called "Snapchat Dysmorphia"); Ex. 942 (internal 25 emails reacting to the JAMA article, with Snap personnel stating, "the idea that filters have given us 26 unrealistic expectations of beauty is well-traveled territory"). Internal communications acknowledged 27 Snapchat users developing "Snapchat dysmorphia," a specific form of body dysmorphic disorder linked 28 to the way in which Snapchat's beauty filters alter the appearance of teenage girls. Ex. 829 at 190:1- 143 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 156 of 235 1 191:13; Ex. 942. Snap's head of user research, Ms. Hammerstrom, acknowledged that lenses can cause 2 "major and emotional pain" to users. Ex. 945 at 2854. 3 Snap could have done the right thing and addressed this concern. Instead, it focused on counter- 4 messaging. Ex. 944 (2018 emails discussing the JAMA article on Snapchat dysmorphia and stating that 5 "continuing to not engage w/ media" is ineffective, and that Snap must push a narrative that lenses "aren't 6 just about beautification. They are silly and wacky"); Ex. 943 (Snap must use the "opportunity" to "find 7 ways to push hard on a counter narrative"). When Snapchat finally—and belatedly—conducted its own 8 research in 2020, it obtained comprehensive data that many users found lenses "troubling" and personally 9 experienced "Lens Dysmorphia." Ex. 946 at 2481, 89, 95-96 ("I look whitewashed," "Lenses aren't made 10 for people of color," "the filters actually make me look decent instead of being a horrendous black girl," 11 "makes my face look thinner and lighter," "I love the filters it hides my ugliness [sic]"). Young users 12 overwhelmingly listed beautification filters as the reason they felt they looked best in the Snapchat camera. 13 Id. at 2495-96. Snap's research concluded that "the prevailing mental model among [younger users] 14 suggests they don't know they are supposed to look any other way [than] in the Camera or in a Lens." Ex. 15 945 at 2854. Snap personnel acknowledged, "using a Lens can create or enhance the dysmorphia users are 16 trying to avoid." Ex. 947 at 6231. Snap Product Manager Juliet Shen testified that she identified "articles 17 or other types of studies showing that [skin lightening filters] could be harmful" and raised these concerns 18 with colleagues at Snap, but the filters still exist on the platform. Ex. 948 at 221:4-222:16. 19 Snapchat still pushes beautification filters on its teenage users to this day. Why? Because Snap 20 understands that beautification filters substantially increase engagement. Ex. 949 at 6749 (Snap's Senior 21 Product Manager for Lenses stating beautification lenses "bri[n]gs us +70% of engagement"). Snap has 22 never warned parents or their children that beauty filters contribute to problematic use, social comparison, 23 eating disorders, depression, and/or anxiety. See Ex. 997 ¶¶ 150, 185, 189, 192, 195, 199, 202; Ex. 983 24 ¶¶ 120-21. 25 (j) CSAM Reporting 26 In the Court's Order resolving Defendants' Rule 12 motions, it made clear that each platform's 27 processes for allowing users to report suspected CSAM was not barred by Section 230 or the First 28 Amendment. See MTD Order, 754 F. Supp. 3d at 963. Snapchat was (and is) a cesspool of pornographic 144 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 157 of 235 1 or semi-pornographic content, as even Meta CEO Mark Zuckerberg has privately recognized. Ex. 304 at 2 2387 ("I just logged into Snapchat for the first time in a while…What surprised me is how porny and 3 inappropriate the content they're featuring is in their equivalent of Reels (Spotlight)…I was completely 4 blown away that they haven't faced more scrutiny for this, especially given that their audience skews so 5 much younger."). Much of this content involves minors, and some of it is child sexual abuse material 6 ("CSAM"), which is illegal contraband under federal law. See, e.g., Ex. 801 at 709:21-710:16 (explaining 7 that teens and adolescents exchange nude photos and videos in group chats and Stories); Ex. 920 at 4453 8 ("sharing of nudes was described as the 'never-ending' problem"); Ex. 993 ¶ 404 (quoting Snap internal 9 documents and observing that "20% of girls and 13% of boys aged 14-18 said they had shared a nude 10 image or video" on Snapchat); see also Ex. 853 at 282:21-283:10 (explaining that Snap does not remove 11 pornography from the platform unless it is reported). 12 Sadly, Snap has failed to provide users adequate tools to report CSAM. For most of its existence, 13 Snapchat did not have a dedicated option to report CSAM. See Ex. 1168 at 19 (as of 2020, no CSAM 14 reporting options for Snaps or Stories). Users wishing to report CSAM were required to identify it through 15 some other reporting category. Id. (users may report nudity or sexual content but no option to flag the 16 involvement of a child). Additionally, Snapchat had no options to report a user for sharing CSAM. Id. 17 Even more egregiously, prior to 2023, Snapchat lacked any mechanism through which users could report 18 content shared through its chat function, despite chat having launched in 2014 and even though Snap knew 19 that the most severe forms of child exploitation occurred through chat. Ex. 1066 at 160:18-162:22; see 20 also id. at 167:14-169:22 (acknowledging that most child exploitation occurs in the chat feature); Ex. 960 21 at 6319 (internal report stating, "TLDR; while chat media is reportable, chat text is currently not. We get 22 feedback from customers, parents, legislators, and advocacy groups that abuse happens in chat text that 23 goes unnoticed due to the lack of reporting functionality"). Even after Snap belatedly launched chat 24 reporting, the volumes of new workflows overwhelmed its Trust & Safety operations and the team was 25 unable to address most reports. Ex. 1131 at 9216. In addition, though Snap personnel recognized that child 26 grooming is "the hardest problem on the platform, particularly given the app's ephemerality, Snap did not 27 implement a proactive grooming detection framework until 2022. Ex. 948 at 260:20-261:4. 28 145 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 158 of 235 1 Snap does not review the reports of CSAM it receives. Snapchat's Transparency Report 2 misleadingly states that virtually all reports seeking content removal are resolved within a reasonable time. 3 Ex. 958 at 147:14-150:15; Ex. 1132 at 7225. But it fails to mention that this statistic does not include 4 reports Snap didn't review before the underlying content expired (or disappeared automatically). Ex. 958 5 at 147:14-150:15. Expired, and therefore unreviewed, reports included approximately 46% of reports for 6 sexual content, 20.5% of reports for harassment, and 17.3% of reports involving threats of harm or 7 violence. See Ex. 961 at 3867-68. Again, because for most of Snapchat's existence there was no option to 8 report CSAM as such, it stands to reason that these expired reports included reports concerning that illegal 9 content. Further, Snap almost completely failed to review reports of accounts, including those trafficking 10 in CSAM. Snap personnel testified that for over four years, users were able to report accounts for 11 violations of the terms of service, but these reports were not reviewed. Ex. 948 at 340:18-342:1 (explaining 12 the lack of workflow available to review all or even a majority of the account reports); Ex. 958 at 154:14- 13 17 (agreeing that Snapchat was "incurring substantial risk from not reviewing these reports"). In 2022, 14 96.7% of account reports were left unreviewed. Ex. 957 at 2520 (note explaining "significant gap" in this 15 "critical safety tool" and that "the content reporting menu no longer reflects [Snapchat's] internal policies" 16 or what Snap "report[s] out on the Transparency Report"). 17 (6) Snap's misrepresentations and manipulations 18 As detailed above, Snap went out of its way not to research issues concerning teen mental health— 19 in order to maintain a posture of plausible deniability. But to be clear, Snap did not remain completely 20 passive either. What it chose to do—where it elected to put its resources—was in developing counter- 21 narratives to dispute, minimize, or reframe evidence of harm. As early as January 2017, Snap was 22 developing "Helpful Pushback Narratives" in response to press articles describing Streaks as a "digital 23 obsession." Ex. 964 at 8501-05. These narratives were crafted to counter reporting that quoted therapists 24 detailing that "the makers built into the app a system so you have to check constantly or risk missing out," 25 and describing Streaks and Snapscore as "unethical designs [which] exploit psychological vulnerabilities 26 to influence what users do without realizing it." Id. Rather than rigorously determine if that were true, 27 Snap focused on insisting it wasn't. 28 In August 2018, when the Journal of the American Medical Association published an article on 146 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 159 of 235 1 body dysmorphia and Snapchat, Snap immediately developed "Sales Messaging" to be "used reactively 2 with clients," downplaying "anecdotal evidence from plastic surgeons" and instructing all employees 3 "speaking at a conference, on a panel, or large event about AR" to review the messaging so they could 4 stay on point. Ex. 965. Internal communications reveal the company's dismissive attitude toward the 5 emerging evidence. Snap's senior communications director described "Snapchat Dysmorphia" as "clearly 6 a made up trend story" and complained that "now it's in a JAMA journal and now people will cite JAMA 7 as concrete-sounding evidence for Snapchat dysmorphia." Ex. 966 at 5654. The concern was not whether 8 beautification lenses were causing mental health harm—Snap made no effort to research that issue for two 9 more years (at which point it found out the article was right). See Ex. 827 at 372:13-23, 562:4-9 10 (acknowledging Snap conducted a user survey on lenses but it was not intended to investigate mental 11 health concerns). The concern was whether the term "Snapchat dysmorphia" would damage the 12 company's reputation. 13 Snap's strategy extended beyond reactive messaging to proactive reputation management. Internal 14 documents show that Snap created what it called a safety cross-functional team—but rather than prioritize 15 safety, it focused on "PR and communications for parental perception" over mental health research and 16 interventions. Ex. 967 at 1617; Ex. 830A at 143:6-160:16. When Snap undertook research on parental 17 perceptions of its platform, the express focus was on developing hypotheses that would "point to 18 opportunities for the company to meaningfully improve parents' perception of the platform" rather than 19 implementing safety measures that parents were actually concerned about. Ex. 968 at 9094-9096; Ex. 803 20 at 175:1-177:15. This strategic approach, which Snap internally referred to as "Project Butterfly," 21 involved coordinating marketing, communications, and public relations efforts to manage stakeholder 22 concerns while systematically avoiding meaningful research into whether the platform was causing 23 documented harms and addressing the underlying cause of those harms. See Ex. 830A at 184:17-186:11. 24 In March 2019, Snap compiled a document cataloguing negative press coverage about Instagram's 25 impact on teen mental health and concluded that there was "a slight advantage for Snapchat." Ex. 969 at 26 3846. The document noted that "Instagram was ranked as the worst offender of young people's mental 27 health and well-being" while "Snapchat was ranked second behind Instagram," and quoted studies finding 28 that those who limited their use of social media apps "showed significant reductions in loneliness and 147 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 160 of 235 1 depression." Id. The clear implication was that if Instagram faced greater scrutiny, Snapchat might escape 2 attention, despite Snapchat's own ranking as the second-worst platform for youth mental health. A 3 responsible company would have taken the message that it needed to mitigate the clear negative impact 4 its platform was having on its youngest users. Snapchat did nothing of the sort. 5 This emphasis on narrative management extended to Snap's most senior leadership. In October 6 2019, when CEO Evan Spiegel gave an interview about Streaks and mental health, he stated that "we've 7 studied this obviously extensively," and claimed that using Snapchat "every day is actually very enriching 8 and it's very different than this idea of community competing." Ex. 970. These statements were made 9 despite the fact that Snap had conducted no comprehensive internal research on mental health impacts and 10 Morgan Hammerstrom, Snap's head of user research, had never been asked to study whether the platform 11 affected users' mental health. Ex. 827 at 96:22-97:13. 12 By 2021, Snap had institutionalized this approach through a comprehensive "message book" 13 designed "to provide messaging on key issues for use in media engagements." Ex. 972 at 1370. The 14 message book explicitly instructed employees that "[e]very question is an opportunity to: Reinforce the 15 'camera company' definition by repeating the word 'camera' and the camera messaging." Id. This strategic 16 framing allowed Snap to deflect questions about social media harms by repositioning itself as a technology 17 company focused on visual communication rather than a social media platform subject to the same 18 concerns as its competitors. This is precisely what Snap seeks to do here. See, e.g., ECF No. 238 at 1 19 (Snap counsel characterizing platform as a camera application different from its co-defendants and 20 "primarily used for direct communication between people who already know each other in real life"). 21 Whether or not that is an effective public relations strategy, it fails to stand up to scrutiny and is certainly 22 no excuse for ignoring the very real harms that Snapchat was causing schools and students across the U.S. 23 (7) Snap failed to warn parents, educators, and schools about 24 known mental health risks 25 Spiegel acknowledged under oath that Snap has "an enormous responsibility to keep [the 26 Snapchat] community safe," that "protecting [Snapchat users] is Snap's moral responsibility," that the 27 "safety of minors on Snap is [the company's] highest priority," and that Snap owes teenagers on Snapchat 28 "a heightened standard of care." Ex. 803 at 16:13-18:5, 26:21-28:3. Further, Spiegel agreed that companies 148 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 161 of 235 1 should warn users of known safety risks on a technology platform and take all reasonable measures to 2 protect teenagers on those platforms. Id. at 20:3-21:5. These sentiments were echoed by Snap's senior 3 management, acknowledging that Snapchat's core audience is vulnerable and it therefore has a heightened 4 duty to protect them. See, e.g., Ex. 801 at 148:21-149:3, 521:1-5. Snap's Senior Director of Product 5 Strategy, Nona Yadegar, explained that the company has "a responsibility to educate our community about 6 the risks . . . they can experience using Snapchat," and one way to accomplish this is by providing proper 7 warnings. Ex. 853 at 133:2-14; see id. at 124:23-125:2. 8 Despite all this, Snap witnesses testified that the company has never warned parents about many 9 known dangers that exist on the platform. Ex. 948 at 281:7-282:8. Nor has Snap ever provided a warning 10 to school districts or teachers that their students were being targeted with advertisements while at school. 11 Ex. 807 at 230:7-231:8. As early as November 2012, one of Snap's outside consultants recommended that 12 the company partner with the Parent Teacher Association ("PTA") to educate Snapchat's user base on 13 safety, security, and privacy issues. Ex. 974 at 8656 (presenting to Spiegel on how to "[p]osition Snapchat 14 as a leader in Internet safety"). Snap did not act on this recommendation or use the PTA to warn the public 15 about known harms and risks. 16 Nine years later, in 2021, Snap communicated with PTAs to jointly survey parents about their 17 knowledge regarding safety features on Snapchat. It was clear during this outreach that parents were 18 interested in learning more. See Ex. 975 (explaining that the surveyed parents "didn't know about any" of 19 Snapchat's purported safety features, "but the good news is that they want to help us get the word out and 20 spread awareness to the PTA community"). At least one PTA expressed interest in working with Snap. 21 Ex. 976 at 9617. Yet nothing progressed for the next three years. 22 In 2024, Snap met with the National PTA and agreed to work together to provide resources to 23 parents. Ex. 977. At this time, the representative from the National PTA stated, "it's just bananas to me 24 that we aren't collaborating with Snap on engaging and educating parents, all things considered." Id. at 25 2531. Unfortunately, Snap's goal in collaborating with the National PTA was not to actually warn parents, 26 students, and schools about the risk of Snapchat use—which it has not done to this day. Instead, Snap's 27 goal was to foster additional growth, as it wished to "create opportunities for kids to share what they love 28 about snapchat with their parents/get parents on Snapchat to understand/use/play with features." Id. Even 149 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 162 of 235 1 this self-serving opportunity ultimately fizzled, after Snap's Policy team cited the "questionable" return 2 on investment. Ex. 978 at 2670 (Beauchere email stating that "the ROI was questionable"). 3 Snap worked with other school organizations to create a Guide to Snapchat for educators in 2024, 4 but it failed to include in this Guide any risks associated with Snapchat use in schools or by students. Snap 5 personnel described their goal to create a better experience within school as it relates to Snapchat. Ex. 801 6 at 714:2-25. Although this Guide includes issue-specific resources, it does not warn about the 7 addictiveness of the Snapchat platform or its tendency cause mental health issues, disrupt the classroom, 8 and harm the school environment. See Ex. 1043. 9 As Plaintiffs' expert, Dr. Minnette Drumwright explains, "Snap's actions are a prime example of 10 pseudo corporate social responsibility. Snap 1) ignored and failed to disclose the harmful and negative 11 experiences that youth encounter on its platform, 2) ignored and failed to disclose that it had inadequate 12 age management technology and processes, and 3) made product decisions that exacerbated addiction 13 among youth and failed to disclose or warn of the dangers. Moreover, in assuring teens, parents, the US 14 Senate, and the American people that Snapchat was safe for teens, Snap did the opposite of warning—it 15 falsely reassured people that Snapchat was safe for young users." Ex. 993 ¶ 388. 16 d) Google/YouTube's Conduct 17 (1) YouTube's business model 18 YouTube LLC, a subsidiary of Google LLC (together "YouTube"), is a multibillion-dollar 19 company that profits from user engagement and on-platform advertising. See generally Ex. 700. YouTube 20 makes money through two primary channels: advertising revenue—which necessitates an emphasis on 21 increased watch time—and paid subscriptions. See e.g., Ex. 701 at 1287 (specific product goals tied to 22 "aim of driving paid subscribers"). YouTube has an expressed goal of fulfilling its financial targets by 23 reaching every child in the world every day. Ex. 702 at 154:25-160:24; Ex. 703 at 3. And YouTube is 24 good at doing just that—it boasted about creating the "YouTube Generation," young people who prefer 25 to binge watch YouTube (for 4+ hours a day) instead of watching cable television. See Ex. 704 at 34. 26 YouTube's ability to leverage users to turn a profit is undeniable and astronomical. YouTube's 27 2015 goal was to reach one billion hours watched a day—expecting this would translate to $15 billion in 28 revenue at then-current monetization rates. Ex. 705 at 2, 5. Since then, its profit margins have steadily 150 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 163 of 235 1 increased. In 2020, YouTube's advertising revenue was $19 billion, Ex. 707 at 2522; by 2023, it had 2 skyrocketed to $31 billion. Ex. 700 at 5637; see also Ex. 700 at 5609; Ex. 706 at 2396.YouTube is 3 consistently touted as a "primary" generator of advertising revenue within the Google and Alphabet 4 pantheon. See Ex. 700 at 5630; Ex. 708 at 3081; Ex. 706 at 2416; Ex. 707 at 2523; Ex. 709 at 2655. 5 YouTube's success is built not only on its number of users, but also their frequency of use. The 6 better it can monopolize and hold user attention, the more YouTube stands to gain. Accordingly, YouTube 7 has prioritized user "engagement" and "Watch Time" as goals. Ex. 710 at 3; Ex. 711 at 3. "[E]ngagement" 8 means nothing more than "quantity of consumption" (and increasing engagement just means "driving 9 greater consumption of videos"). Ex. 772 at 7889. Likewise, Watch Time is straightforwardly defined as 10 "How much time the user spends watching videos." Ex. 713 at 1686. Watch Time "has a long and famous 11 history at YouTube" as a key metric. Id.; see Ex. 710 at 3 (2017 growth team presentation acknowledging 12 that total watch time has always been a key metric); Ex. 712 at 4105 ("primary metric that YouTube's 13 Search and Discovery systems optimizes around is time"); Ex. 985 ¶ 130. 14 YouTube's goal of increasing revenue by driving more engagement and Watch Time resulted in 15 two areas of focus, detailed below. First, YouTube targeted young users—the younger the user, the longer 16 the potential lifespan of engagement and, in turn, revenue stream. YouTube specifically targeted 17 educational spaces, seeing an opportunity to engage a captive audience, build early brand loyalty, onboard 18 kids into the broader Google ecosystem, and dominate young lives during their likely first interactions 19 with technology outside the home. See Ex. 740 at 148:13-149:5; Ex. 741 at 10, 18. Second, YouTube 20 designed for addiction—knowing that sustaining user engagement and attention required features that 21 would cause users to return more and more often. Among other design choices, YouTube aims to fill even 22 small moments in young people's lives with watch time, using push notifications to deliver "snackable" 23 content (30-second to minute long short-form videos) for rapid consumption, while users wait for the bus 24 or stand in line. Ex. 715 at 380:9-19; id. at 523:16-524:5. YouTube did this without any warning to the 25 public or school districts—despite knowing that "[d]riving more frequent daily usage [was] not well- 26 aligned with...efforts to improve digital well-being." Ex. 715 at 389:9-20; Ex. 772 at 7900. 27 (2) YouTube's targeting of school-aged children 28 According to its witnesses, YouTube is not a social media platform at all, Ex. 1012 at 232:7-9, but 151 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 164 of 235 1 a mere "video distribution platform." Ex. 1003 at 200:25-201:3; see Ex. 1001 at 280:14-17 ("[Y]ouTube 2 is not a social media app. We don't have a social graph."); Ex. 1002 at 330:1-10; 343:11-13. Regardless 3 of what it is called, the reality is that YouTube competes fiercely with its co-defendants in the race to 4 attract and hold the attention of young users. See Ex. 772 at 7898 (discussing competition with Facebook 5 and Instagram's feed-based consumption model which allows for ad scaling and revenue growth)) Ex. 6 701 at 1275, 77 (comparing design features of Snapchat and Instagram to YouTube). And, by many 7 measures, YouTube is winning that race. Through a combination of ease of access, addictive features, and 8 embedding in the school environment, YouTube has become the most popular platform for teens. A 2024 9 Pew Research Center study concluded that "73% of teens say they go on YouTube daily, making YouTube 10 the most widely used and visited platform we asked about. This share includes 15% who describe their 11 use as 'almost constant.'" Ex. 1038 at 2 (2024 Pew study comparing social media platforms YouTube, 12 TikTok, Instagram, and Snapchat (emphasis in original)). 13 YouTube has dedicated significant resources to growing its youth user base, which it knew could 14 be a rich and enduring source of user engagement and profit. "If we want to be successful with teens, 15 besides building features to retain them once they arrive on [YouTube], we also need to focus on the 16 experience for the next generation of teens—today's younger kids and tweens." Ex. 724 at 5689. "This is 17 an existential question for [YouTube]—We need to make big bold bets in this space if we want to ensure 18 that we keep these users on the platform to become teens on [YouTube]." Id.; see also Ex. 725 at 13 19 (Investment in youth is a "business opportunity for YouTube" to "retain and grow the next generation of 20 creators, otherwise they could leave for other platforms and never come back"); see also Ex. 726 at 9331 21 ("Aging-up is of strategic importance to YouTube overall…Thus, defining a cogent strategy to address 22 the 8-13 market is a key challenge and opportunity for YouTube"). 23 To that end, YouTube developed a suite of products that would indoctrinate children early into the 24 YouTube experience and create a seamless product progression that follows them as they grow up. In 25 2015, YouTube launched YouTube Kids, which it designed to attract children under the age of 13. See 26 Ex. 730 at 1; Ex. 703 at 3-4 ("YTK Vision: Create the next generation of YouTube Users" and make 27 "YouTube (Kids) as a part of the daily routine for all kids under 13") (emphasis in original); Ex. 727 at 7 28 (target audience is under 13 year olds). In designing YouTube Kids, YouTube identified sub-age groups 152 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 165 of 235 1 to target with specific content, with categories of users ages 8-12, 5-7, and even 4 and under. Id. As one 2 executive succinctly framed it: "is there strategic value to having a kids product at 3 [YouTube][?]…yes…they grow up[.]" Ex. 728 at 1442; id. at 1443 ("we have [YouTube] kids for them 4 to shed as their 'fisher price' moment"). YouTube Kids was thus understood to be a gateway to use of the 5 main YouTube platform. 6 Despite targeting YouTube Kids at children as young as four years old, YouTube made seemingly 7 little effort to understand the developmental impacts of this platform. Shimrit Ben-Yair, the Project 8 Manager for YouTube Kids at its inception, denied having discussions with experts regarding children 9 and screen time (aside from speaking to her own children's pediatrician, who advised that all screen time 10 be limited to two hours a day). Ex. 1051 at 248:12-25. To the contrary, Ben-Yair admitted that "a lot of 11 our initial work on YouTube Kids was guided by our own experience as a parent, and as parents," id. at 12 106:3-106:21—not exactly the kind of rigorous diligence one would expect from a Fortune 10 company. 13 Indeed, this slapdash, anecdotal approach led to design decisions that actually contradicted what little 14 research YouTube undertook. YouTube Kids was initially available only on tablets. Id. at 76:16-24.,105:2- 15 10. Ben-Yair testified that this decision was guided by hearing that tablets were families' preferred method 16 to gather and watch videos. Id. In actuality, YouTube's official research found that television is still most 17 often used for kids 5 to 7, bigger screens are better for parental "earshot monitoring," and educational 18 media time drops by 16 percent when kids are on mobile devices (perhaps because "young kids are often 19 holding the tablet" when it is made available). Id. at 96:9-106:2; Ex. 1006 at 5-6, 11, 43-45. 20 Not surprisingly given its origin story, YouTube Kids was and is ineffective as a supervision tool. 21 First, as detailed more below, any child can continue to access the main YouTube platform, either without 22 an account or with one (which is easy for any child to create). See § III.A.2.d.5.d. Second, the supervision 23 aspects of YouTube Kids were woefully deficient in terms of addressing addictive and compulsive use. 24 Parental controls for YouTube Kids centered on content moderation—limiting the corpus of content 25 available, blocking certain content, and disabling search. See Ex. 730 at 3. While YouTube Kids had a 26 built-in timer to set screen time limits (see id.), it also retained many addictive features, like autoplay, 27 notifications, recommendation systems, and scrolling. 28 153 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 166 of 235 1 In fact, when YouTube launched autoplay on YouTube Kids it did so without a "toggle," i.e., 2 without any ability for anyone—parent or child—to turn it off. See Ex. 789 at 1972 (2019 document 3 discussing the plan to add an on/off toggle). Ben-Yair's only justification for this design decision was 4 that, "if you as a parent said that you want your kids to watch 15 minutes of YouTube, but then every two 5 minutes you had to go play the next movie for them, then that would not be the experience that parents 6 wanted." Ex. 1051 at 91:5-16. In reality, what YouTube's research showed is that parents wanted 7 control—the choice to make decisions for themselves and their children, including whether or not such a 8 feature should be enabled. Ex. 1006 at 5-6, 11, 43-45. It was not until 2022—seven years after its launch— 9 that YouTube created a toggle (and set autoplay to default off). See Ex. 1031 at 38:1-39:2. 10 YouTube Kids was not the company's only effort to target an under 13 year old audience. After 11 its launch, YouTube recognized that it had a product gap for tweens—and needed a way to retain those 12 users, or risk losing them to other platforms. See Ex. 731 at 6490 (raising the alarm that "[t]here is no 13 path today for kids to participate/create on YT…users are trying to satisfy this demand elsewhere," 14 citing TikTok as an example (emphasis in original)). So, in 2021, YouTube launched SupeX, an 15 experience designed for tweens that was intended as a bridge between YouTube Kids and YouTube Main. 16 See Ex. 729 at 6 (discussing YouTube's retention gaps); see also Ex. 730 at 3 (SupeX "designed for tweens 17 transitioning to YouTube Main"); Ex. 731 at 6489 ("SupeX has the content missing from [YouTube 18 Kids]…and would satisfy this age group"). Like YouTube Kids, SupeX is deficient as a supervision tool, 19 as discussed in more detail below. See infra § III.A.2.d.h. 20 In the hunt for young users, YouTube also began to emulate features of other social media 21 platforms. By 2018, YouTube determined that it was "the most widely used internet Platform by US teens 22 (who watch an average of 1 hour and 22 min / day)." Ex. 772 at 7914. But this wasn't enough—YouTube 23 was "behind Snapchat in the peer-to-peer space and Instagram in the public short form video space," so it 24 scrambled to figure out what features it could offer to remain competitive. Ex. 701 at 1275, 77. In the hunt 25 for young users, YouTube began to emulate features of other social media platforms. YouTube launched 26 "YouTube Shorts" in 2020, which mimicked the vertical, short-form video format of TikTok's For You 27 Page and Instagram Reels. See Ex. 1052 at 229:20-230:10. YouTube did so because it knew that this 28 format was popular with young users. See Ex. 733 at 8317 ("Shorts is our big thing for teen appeal"). In 154 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 167 of 235 1 tandem with Shorts, YouTube employed notifications to maximize Shorts visits, sending users messages 2 about Shorts at what it perceived to be moments ripe for "snackable consumption," i.e., watching content 3 that can be viewed in little bits. Ex. 715 at 380:9-19; id. at 523:16-524:5. 4 YouTube also sought to capitalize on the popularity of effects/filters, including "Appearance 5 effects" (defined as "all effects that alter someone's appearance"). Ex. 1071 at 6. YouTube knew these 6 features resonated with young users on competitor social media platforms. See Ex. 734 at 6752 (exploring 7 opportunity to engage users with effects and noting that Snap, Meta, and TikTok have over 2 million 8 effects with over 100 million daily active users). It also knew that "expert researchers found beauty filters 9 and negative social comparison may be correlated with negative wellbeing, especially in youth" and "that 10 the pervasive nature of filtered images regularly trigger body dysmorphia." Ex. 1074 at 7. 11 (3) YouTube's targeting of schools 12 YouTube knew that it was changing the behavior of a generation—in which 53% of teens said 13 they "would drop what they're doing to watch YouTube," 68% were aspiring influencers, and 45% 14 thought it was important to define their online personalities. Ex. 704 at 28, 41. But even this wasn't 15 enough. YouTube wanted the attention of young people during the school day. For more than a decade, 16 YouTube has relentlessly pursued school adoption and student use during the school day, launching 17 numerous platforms and tools aimed at increasing YouTube's school presence and encouraging use by 18 students outside of the classroom. By 2013, YouTube was examining the "smartphone / tablet / app 19 explosion" in "Pre-K / Primary," and investigating the possibility of "flip[ping] the classroom" in high 20 school, see Ex. 752 at 24—by which teachers would replace in-class lectures with YouTube videos 21 assigned as homework. Ex. 752 at 24-25. 22 YouTube's widespread use in schools is no accident. Google seeks to get kids using the Google 23 "ecosystem"—apps and devices—early and extensively. Ex. 741 at 9, 14, 16; Ex. 1135 at 1, 2, 7, 11. Kids 24 are a "Massive Opportunity" for Google because early use builds "brand trust and loyalty" that impacts 25 kids' choices throughout their "lifetime"—particularly as to which apps to use. Ex. 741 at 1, 5, 9, 16, 17. 26 Google quickly realized that the key to unlocking this opportunity is getting kids using the Google 27 ecosystem at school. Id. at 17 ("Investing in schools helps onboard kids into Google's ecosystem."). 28 YouTube saw two key benefits from infiltrating schools. First, doing so would increase profits via 155 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 168 of 235 1 increased watch time of educational content (i.e., increased ad revenue) and/or increased paid 2 subscriptions for its education-tailored services. Ex. 742 at 4; Ex. 752 at 9, 17. Second, infiltrating schools 3 would establish a user "pipeline" to ensure the youngest generation was brand-loyal to as many Google 4 products as possible (Chromebooks, G-Suite, YouTube, etc.). Ex. 740 at 148:13-149:5; Ex. 741 at 10, 18 5 (explaining that onboarding kids leads to brand trust and loyalty over lifetime and school laptop brand 6 influences future purchase patterns); Ex. 742 at 4 (noting the opportunity to create a pipeline of future 7 users and creators); Ex. 740 at 193:2-195:11. 8 Google's focus on kids in schools was unprecedented. Using a fit-for-purpose "K-12 Sales Team," 9 Google "took over" K-12 learning in the U.S. in about five years, pushing aside other tech giants like 10 Apple and Microsoft. Ex. 741 at 16 (citing "Google Took Over Classroom" case study), id. at 49 ("Google 11 has won the K-12 education technology market"); see generally Ex. 1136 ("North America K-12 EDU 12 Sales Team Overview"). The key to Google's "ecosystem play" in U.S. schools were its low-cost Google 13 Chromebook and its free apps for teachers and kids to use—including YouTube. Ex. 741 at 17-18 14 ("Ecosystem play" based on which laptop brand used and "[e]ven without Google hardware, kids are still 15 using Google services."). Chromebooks could be used by kids in classrooms to watch videos on YouTube, 16 Ex. 740 at 158:14-24, and YouTube in school was another way to embed kids into the Google ecosystem. 17 Id. at 148:13-149:5. Indeed, "video" was one of the pillars of Google's effort to "win in the EDU space" 18 and the K-12 Sales Team saw itself as providing a "bridg[e]" to "all of Google," including "YouTube." 19 Ex. 1135 at 2; Ex. 1136 at 2. 20 Having inserted itself into the classroom environment, YouTube knew that it needed some way to 21 normalize that state of affairs to teachers and principals. Over the course of more than a decade, YouTube 22 has rolled out various palliations, which (until well after these lawsuits were filed) have been largely 23 ineffective and designed as much with YouTube's business objectives in mind. These platforms and tools 24 include: (1) YouTube for Schools/EDU-mode (2011), Ex. 743 at 4725; Managed Restricted Mode 25 ("MRM") (2015), Ex. 745 at 25; Ex. 744 at 9512; Player for Learning (conceptualized in 2020 but 26 launched as Player for Education in 2023), Ex. 746 at 7; Ex. 747 at 11; Ex. 749 at 2038; and school time 27 controls (2024), Ex. 1002 at 205:15-206:10; Ex. 766 at 470:2-20. 28 156 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 169 of 235 1 YouTube's first attempt to make its platform more palatable to schools launched in October 2011 2 and allowed schools to sign up for what it called YouTube for Schools. Ex. 1076 at 9257; Ex. 743 at 4725. 3 YouTube for Schools, which launched the platform in what YouTube called "EDU-mode," was a selection 4 of videos YouTube determined were "appropriate for educators." Ex. 743 at 4725; Ex. 740 at 5-15. EDU- 5 mode restricted which videos could be played and disabled other functionalities. Ex. 743 at 4725. 6 YouTube also funded EDU channels, which are channels offering academic videos related to subjects 7 taught in school or videos related to digital, professional, and trade skills. Ex. 1077 at 4; Ex. 752 at 12. 8 Internally, YouTube was not coy about its motivations for prioritizing EDU-mode. YouTube 9 imagined a world where "Parents ask their children 'Why aren't you watching more YouTube?'" and 10 "School Administrators shift budgets from Textbooks to YouTube subscriptions." Ex. 752 at 10. It 11 understood that "EDU has [the] opportunity to be [a] key driver of view growth" and that "More EDU 12 content = More watch time per user." Ex. 752 at 9, 17; Ex. 740 at 126:10-131:24; see also generally Ex. 13 753 (2015 internal memo identifying "YouTube EDU Opportunities"). Indeed, in the aggregate, 14 YouTube's EDU channels garnered over 3.5 million hours (400 years) of watch time. See Ex. 752 at 13. 15 Goals aside, YouTube for Schools was known by YouTube to be a complex technical solution that 16 was easily circumvented by students. Ex. 743 at 4725. The solution required schools to sign up and then 17 configure their network to send all outbound HTTP traffic to YouTube with an HTTP header or URL 18 parameter identifying their school. Ex. 1076 at 9257. Once this was done, in theory schools had access to 19 the YouTube for Schools global whitelist of 1800 "EDU Channels" as well as the option for admin-created 20 playlists. Id. But in practice, the sign-up flow was broken and YouTube for Schools experienced "lots of 21 outages." Ex. 1076 at 9257. (This was not surprising given that YouTube for Schools was a consistently 22 underfunded project (referred to internally as a "perpetual 20% project"). Ex. 745 at 73.) As a result, 23 YouTube for Schools was not widely adopted. Ex. 743 at 4725. Ex. 744 at 9512. In 2012, YouTube for 24 Schools had only been adopted by 4% of approximately 14,000 school districts in the United States. See 25 Ex. 750 at 5; Ex. 751 at 5. Further, in September of 2012, EDU-mode was rendered toothless when 26 YouTube became accessible via the HTTPS protocol (https://www.youtube.com). This allowed students 27 to easily sidestep EDU-mode's protective filter. See Ex. 743 at 4725; see also Ex. 750 at 5; Ex. 751 at 5. 28 In the summer of 2015, YouTube launched Managed Restricted Mode ("MRM"), advertising that 157 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 170 of 235 1 it would allow school administrators to select one of three access settings and, in doing so, limit the corpus 2 of videos available to students accessing YouTube on their school network. Ex. 1078 at 85 ("admin sets 3 MRM controls at the network-level (e.g. any computer on the school network is categorized as MRM"); 4 Ex. 745 at 25 (depicting permission levels available to school implementing MRM). Google Classroom 5 users also had the option to set access restrictions for managed Google accounts. Ex. 1078 at 85. But mere 6 months after launch, YouTube internally acknowledged that MRM was "not meeting school's needs," Ex. 7 753 at 4419, because its "very basic filtering options" did not prevent students from receiving, uploading, 8 or sharing non-educational videos or watching distracting ads. Ex. 753 at 4419; Ex. 740 at 132:2-133:2, 9 190:13-191:9. YouTube knew that, despite the importance of MRM filtering to school districts, its MRM 10 safety features were "trivially easy" for students to bypass. Ex. 1011 at 1769.ECM (listing several ways 11 students could avoid protections, such as 12 . In addition, the descriptions for the access settings (Strict Restricted, Moderate 13 Restricted, and Unrestricted) were "too vague to be applied," causing frustration for administrators who 14 were unable to explain to teachers and staff why certain videos were blocked. Ex. 1078 at 10. (Based on 15 the date of this document in 2021, this problem remained an issue six years after MRM launched.) Finally, 16 MRM did nothing to limit students' use of YouTube during the school day on personal smartphones over 17 cellular networks. 18 YouTube also knew it was not a proven platform in the classroom context. Ex. 757 at 9082 ("The 19 YouTube experience in K-12 schools is broken."); Ex. 740 at 60:23-61:1 (YouTube has no data showing 20 that use of YouTube by K-12 students results in better grades); Ex. 758 at 36 (YouTube "isn't designed 21 for formal learning."); Ex. 745 at 31 (admitting MRM was only meeting schools' "baseline need" and did 22 not work with other tools like Classroom and Chromebooks); Ex. 740 at 262:1-263:12; Ex. 757 at 9014 23 (acknowledging MRM has been "historically under-resourced," resulting in "increasingly common user 24 issues and emerging business risks."). 25 Indeed, YouTube recognized that its disorganized library of videos, along with the high potential 26 for users to be distracted from intentional viewing of educational content, made YouTube problematic in 27 the educational context and limited its ability to deliver structured learning. Ex. 740 at 169:15-170:19. A 28 key problem is YouTube's "Watch Next" feature—which would routinely offer "recommendations [that] 158 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 171 of 235 1 don't align to a defined learning journey," making it hard for students to "stay on track." Ex. 755 at 7404. 2 One internal presentation provided an example, showing a user watching a YouTube video about linear 3 equations and then being recommended a "Will Ferrell hilarious acceptance speech." Ex. 745 at 74. As 4 the presentation acknowledged: "Hard to do my homework when I could be watching Will Ferrell." Id. 5 But YouTube also knew that allowing users to remain focused on their original learning intention meant 6 fewer views for non-learning videos that YouTube might recommend—meaning less engagement and less 7 Watch Time. Ex. 755 at 7404 ("If you remove non-learning things from the watch next of learning videos, 8 then those videos will have far less reach."). 9 Despite knowing that its use case as an educational tool was unproven, and despite knowing its 10 existing safety measures for schools and students were ineffective, YouTube embarked on a campaign to 11 increase school-day and classroom usage even more. Specifically, YouTube set out to close what it 12 perceived to be an 80 million hour difference in watch time between any given Thursday and any given 13 Saturday. Ex. 753 at 4419. YouTube understood this to be an engagement frontier for the platform and a 14 business "opportunity in education." Ex. 740 at 129:25–133:7-15. In furtherance of its goals, YouTube 15 pursued partnerships with organizations like the National Parent-Teacher Associations (PTA) and 16 Common Sense Media due to their reputation as highly credible sources of information for educators and 17 parents. See Ex. 1002 at 250:5-251:5, id. at 71:8-11. YouTube then used those partnerships to distribute 18 information to schools and normalize itself as part of the classroom experience. 19 YouTube's agreement with the PTA was a paid partnership. Ex. 760 at 6433 (in 2022-23, 20 for basic renewal plus + to add PTA grants and conference fees). The partnership provided 21 YouTube with a direct line of communication to PTA members. In 2022, for instance, YouTube 22 employees presented on their "efforts for younger kids, tweens, teens and parents" at an hour-long session 23 at the annual PTA convention. Ex. 762 at 3364. YouTube also reached PTA members across the nation 24 through email lists and editorials in the PTA newsletter, such as one about YouTube's Youth Principles 25 drafted by James Beser, Senior Director of Product Management for Kids & Youth. Ex. 761 at 0318. 26 While these communications described the dangers of various third-party content (for instance, "viral 27 hoaxes") they conspicuously excluded information about potential harms presented by YouTube itself. 28 See Ex. 1002 at 100:2-24, 111:7-25; Ex. 740 at 349:25-350:18. 159 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 172 of 235 1 Like the PTA, Common Sense Media is a non-profit organization that is well-regarded by parents 2 and educators. Ex. 1014 at 7, 20. Its entire focus is children's media, Ex. 1002 at 261:17-20, and it serves 3 as a clearinghouse for safe technology and media guidance for kids, Ex. 1014 at 17-19. Among other 4 functions, Common Sense Media releases movie and TV ratings to assess what content is kid-appropriate, 5 id. at 263:18-264:8, and provides digital literacy curriculums for K-12 students. Ex. 1079 at 2. 6 The high-level goal of YouTube's partnership with Common Sense Media (CSM) was to drive 7 improved trust among parents and key opinion formers ("KOFs"), Ex. 1080 at cell C13—including 8 regulators, Ex. 1001 at 183:22-184:2, and legislators, Ex. 1024 at 227:9-17. YouTube recognized that 9 consumers are more persuaded by authentic, unsponsored feedback—an area where CSM excelled. See 10 Ex. 1002 at 254:16-255:7. YouTube also understood that parents in particular react well to seeing CSM's 11 green check mark and branding, which feels vetted and safe. Ex. 1014 at 7, 20. By partnering with CSM, 12 YouTube positioned itself as a platform safe for use by children, including children under the age of 13. 13 Among other promotions, the partnership resulted in two youth content creation guides (one for teens and 14 one for parents) and sponsored content on YouTube-branded websites directed toward parents. Ex. 1080 15 at C14. Notably, despite a stated objective of "educat[ing] stakeholders and users," id., none of this 16 material refers to the risks of addiction and compulsive use that YouTube knew are inherent to its platform. 17 In December 2022, CSM and Frances Haugen co-authored an amicus brief filed in Gonzalez v. 18 Google which, among other things, argued that "Section 230 does not grant blanket immunity" for a 19 variety of Google's "non-publishing activities." Brief for Common Sense Media and Frances Haugen as 20 Amici Curiae Supporting Petitioners at 3, Gonzalez, et al., v. Google LLC, No. 21-1333 (Dec. 6, 2022). 21 Months later, CSM published a report on Teens and Pornography that was highlighted by the New York 22 Times. Ex. 886 at 20, 22; Ex. 888 at 1. In response to these developments, Google prepared a dossier 23 regarding its numerous business and financial entanglements with CSM and asked for a meeting to discuss 24 their deals—including a partnership with CSM's for-profit subsidiary, Common Sense 25 Networks, and a Google TV partnership which cost Google annually. Ex. 1111 at 0511-12 26 (talking points: "In June 2022 CSM did not hesitate to call Google out publicly, despite our active 27 partnership deal."). YouTube's message to CSM was not particularly veiled: "We've heard feedback from 28 160 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 173 of 235 1 the teams that while they are not against working with Common Sense Media, there is hesitation around 2 the scale, fit, and cost of the products." Id. at 0512. 3 Ultimately, YouTube's brand cleansing worked, and YouTube's incursion into schools was a 4 success. Today, YouTube's CEO, Neal Mohan, is confident that more than 90 percent of teachers use 5 YouTube in the classroom every single day. Ex. 1012 at 168:21-23. And still, YouTube has no data 6 showing that the use of YouTube by K-12 students improves reading skills or any other skill, as YouTube 7 does not measure the effectiveness of its efforts with respect to the use of YouTube in classrooms, Ex. 8 740 at 61:16-20; 64:24-64:6. In fact, what YouTube measures is users "really valu[ing] YouTube video"— 9 YouTube's euphemism for engagement with the platform. Id. at 64:7-65:18; see Ex. 772 at 7889 10 ("satisfaction" still measured in part by Watch Time). 11 In 2022, after more than a decade of pushing its platform into the classroom space, YouTube 12 launched a product called Player for Education. Ex. 740 at 43:17-44:5, 51:14-52:4, 162:20-163:15; see 13 also Ex. 749 at 2037 ("full rollout" of Player for Education in mid-2023); Ex. 746 at 7. Player for 14 Education enables teachers to assign specific YouTube videos to students via an embedded player within 15 Google Classroom, or other Google products, with no links back to YouTube, no recommendations, no 16 ads, and no cookies.7 See Ex. 740 at 43:17-44:5. This solution attempted to address several problems 17 reported by educators, while removing impediments to greater adoption of YouTube in classrooms. 18 However, Player for Education does nothing to address the issue of student access to YouTube during 19 school hours or on personal devices. 20 Finally, YouTube delayed implementing any screen time management controls effective during 21 the school day until late 2024 (well after these lawsuits were initiated). Before then, granular school-day 22 controls were not available, and parents had to attempt to devise workaround methods to limit use of 23 YouTube during school hours. For example, in March 2021, YouTube employees noted in an internal 24 7 25 YouTube's 30(b)(6) corporate representative on school district issues, Katie Kurtz, briefly mentioned an embed player pre-dating Player for Education. See Ex. 740 at 52:5-13. YouTube's iteration of the embed 26 player that preceded the Player for Education was never a real solution for teachers because it was hard to use and, even when used, still showed "related videos." See, e.g., Ex. 931 at 9197 (internal chat: "teacher 27 must embed in a specific way" that even led Google staff to get "An error occurred' message); Ex. 1174 28 at 3620 (embed player used in Google Classroom played "related videos"). 161 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 174 of 235 1 discussion that they were using the bedtime settings to block their children's YouTube access during the 2 school day. See Ex. 766 at 476:21-477:8; 477:15-480:3; Ex. 764. YouTube did not launch a school time 3 feature, to assist parents with controlling their children's YouTube access during school hours, until 4 August 2024. See Ex. 766 at 470:2-20; Ex. 1002 at 205:15-206:10. 5 (4) YouTube's knowledge of harms 6 YouTube acknowledges that "kids [and] teens" are "vulnerable groups." Ex. 736 at 33. It knows 7 that "brain development predisposes young teens to act more impulsively [and] show great tendency 8 towards risk taking" and that "[e]xecutive functions aren't fully developed at this age (e.g., self-regulation 9 and decision-making capabilities)." Ex. 735 at 5; see Ex. 1010 at 156:3-16; Ex. 737 (YouTube CEO: 10 "[t]eenage brains are still developing and work differently than adult brains, particularly when it comes to 11 decision-making or self-regulating"); Ex. 738 at 7 (identifying vulnerabilities specific to teens); Ex. 739 12 at 38 (citing a consistent increase of self-harm and suicide rates among adolescents and noting "Teens are 13 more vulnerable because their judgment and decision-making abilities are still being developed."). 14 Despite this knowledge, YouTube designed its platform in a way it knows impacts youth wellbeing 15 and promotes addictive and compulsive use. A 2016 presentation to executives admits as much: the 16 YouTube Main App team "aspire[d] to create an app that is…Addictive[:] Our app experience should 17 compel users to come back more and more often." Ex. 765 at 11. (In this same presentation, YouTube 18 reported that it was reaching over six billion daily active views in the U.S. alone. See id. at 14.) Other 19 presentations are similarly candid. See also Ex. 701 at 1255 (goal to "[i]ncrease habitual users" and "focus 20 on making YouTube a daily habit"); id. at 1235 (discussing "aim of increasing frequency and long-term 21 YouTube time"); Ex. 767 at 5098-5099 ("Watch time, and only watch time;" "All other things being equal, 22 our goal is to increase (video) watch time"); Ex. 768 at 6246 (identifying notification features to "help 23 build a YouTube habit[]"); Ex. 769 at 2 (presentation by YouTube Growth Team exploring "Principles of 24 Habit Building" and brainstorming "features and concepts that build habits on YouTube"). 25 By at least 2018, YouTube knew that "tech addiction" had become a "major topic of concern," 26 particularly the risks it presented "to the physical, emotional, and social development of teens." Ex. 772 27 at 7914. It identified three areas of concern impacting users 13-24 disproportionately: 28 162 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 175 of 235 1 2 3 4 5 6 7 8 Id. Despite gaining an awareness of widespread "habitual heavy use," "unintentional use," and sleep- 9 disruptive "late night use" by teens, id., YouTube did nothing to warn the public or the Plaintiffs. Instead— 10 within the very same document—YouTube focused on Meta's ability to "scale up monetization faster than 11 YouTube." Id. at 7898. YouTube proposed solving this "problem," first, by continuing to "optimize for 12 getting people to initiate long video watching sessions"; second, by introducing new feeds to promote 13 "more scrolling," id.; and third, by developing a video format "optimized for the phone." Id. at 7960. 14 In 2018, YouTube conducted various studies and literature reviews that shed further light on the 15 scope of the addiction problem. One of these reviews noted: "Research suggests that a common side-effect 16 of using platforms with on-demand video can result in binge-watching and procrastination (Pena, 2015)." 17 Ex. 773 at 19. "It is also known that YouTube's algorithms are designed in a way to increase 'rabbit hole' 18 watch time and 'keep people hooked on the screen' (Lewis, 2018)." Id. at 19 (emphasis removed). Another 19 defined "Technology Addiction" as the "widespread obsession, compulsive and excessive use of 20 technology that interferes with Daily Life"; acknowledged research "indicat[ing] that [the] prevalence of 21 tech addiction is inversely related to quality of life"; and observed that "Tech Addiction and Google's role 22 has been making the news." Ex. 774 at 3. This same review flagged YouTube's autoplay feature as 23 "Compulsive Content Consumption" and labeled it a disrupter of sleep patterns. Id. at 6, 8. 24 Still another literature review acknowledged that "excessive video watching is related to 25 addiction"—and concluded that watching short-form videos gives users a "'quick fix' of dopamine" that 26 is "[s]imilar to feelings of reward when using drugs or other similar substances." 27 28 163 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 176 of 235 1 2 3 4 5 6 7 8 9 10 11 12 Ex. 775 at 13; see also Ex.1081 at 290 (acknowledging that short-form content "exacerbates concerns of 13 addiction and ADHD"); Ex. 1082 at 5298. (Trust & Safety employee asking whether YouTube has "any 14 responsibility to moderate binge watching"). Despite its understanding that short-form videos can trigger 15 an "addiction cycle," Ex. 775 at 4191, YouTube proceeded to develop Shorts (its short-form video format), 16 which it released in the United States approximately three years later. Ex. 1112 at 6. 17 YouTube's negative effects on millions of young people—and the panoply of harms it could have 18 and should have warned the public about—are laid bare in a presentation titled Double Rainbow: A Vision 19 of Digital Wellbeing at YouTube. Ex. 738 at 2. In this document, which employees authored in 2018 but 20 iterated on for years, YouTube acknowledged its negative effects on wellbeing, including "Sleep 21 quality/duration," "FOMO," "Isolation," "Depression," "Anxiety," and "Body/Esteem issues." See id. at 22 15-16, 28 (evidencing comment from 2018). It admitted: "YouTube is designed around increasing users' 23 engagement, not maintaining user's intention." Id. at 73. It observed: "Experts compare YouTube to a slot 24 machine, with variable rewards and infinite possibilities." Id. at 77. In another version of Double Rainbow, 25 YouTube concluded, based on user survey evidence, that 10% of users aged 13-24 watched 2 hours a day 26 habitually, Ex. 1028 at 41, and 5% watched 3 hours a day habitually. Id. One 16 year old reported, "It 27 delayed my sleep a lot. I'd watch until 2:00 or 3:00 in the morning...I'd be so invested that I'd keep 28 watching." Id. at 9. 164 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 177 of 235 1 This same document estimated that 69 million teens were on YouTube. See id. at 41. 2 (5) YouTube's failure to exercise reasonable care 3 YouTube had an obligation to ensure that the YouTube Platform was appropriate and safe for 4 young users—a demographic it was intentionally targeting. See, e.g., Ex. 1013 at 27:3-7 ("if we learn 5 about something that has kind of a cause-and-effect impact to youth, we would take action to…adjust our 6 product experiences to make sure that it is most appropriate for younger users"); Ex. 714 at 45:16-46:3, 7 53:23-55:22 (YouTube's former Senior Director of Trust and Safety agreeing that companies should 8 assess products for risks to users, that this obligation is ongoing, and that YouTube has a responsibility to 9 ensure that its products are safe for children, and to warn parents and the public of any risks). In 2024, 10 YouTube even codified some of these principles in writing, authoring a "Youth Legislative Framework" 11 that would require online services to prioritize the best interest of children and teens in the design of their 12 products. Ex. 770 at 2; see generally Ex. 771. YouTube failed to live up to these principles or obligations. 13 To the contrary, it knowingly embedded a bevy of harmful, addictive features in the YouTube platform. 14 (a) Autoplay 15 Autoplay was launched on YouTube in 2015. See Ex. 789 at 1950. It is designed to create a 16 seamless viewing experience, by automatically and continuously playing one video after another without 17 any affirmative user action. See id. At the time of its launch, and up until 2021, the default autoplay setting 18 for all users, regardless of age, was "on." See id. at 1952. This was a business decision, made without any 19 concern for user wellbeing: "We expected Autoplay on by default to generate more WT [Watch 20 Time]." Id. (emphasis in original); see also Ex. 790 at 9 ("More autoplay = more WT, and that was an 21 explicit goal of the project"). YouTube was right. Shortly after launch on desktop (2015), autoplay was 22 credited with generating 16% of YouTube's desktop watch time. See Ex. 791 at 3300. Its growth 23 contribution continued through 2019, when it was credited with driving of total Watch Time across 24 all interfaces. See Ex. 789 at 1954. YouTube has internally lauded Autoplay as "the single most impactful 25 launch in YouTube history." See Ex. 765 at 28. 26 YouTube knew that autoplay contributed to excessive and addictive use. See, e.g., Ex. 789 at 1962 27 (acknowledging autoplay creates wellbeing issues for some users, which includes prolonged unintentional 28 165 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 178 of 235 1 use and sleep disruptions); Ex. 790 at 10 ("Autoplay's share of watch time doubles at night"); see id. at 2 12 (acknowledging concerns about autoplay from media, key opinion formers, and legislators); Ex. 775 3 at 13 ("Researchers feel that YT is built with the intention of being addictive [and] [d]esigned with tricks 4 to encourage binge-watching (i.e. autoplay…)"); see also Ex. 987 ¶¶ 59-61, 200-202; Ex. 982 ¶ 193; Ex. 5 1004 ¶ 445; Ex. 983 ¶ 5.b. Nonetheless, it failed to inform its users, their caretakers, Plaintiffs, or the 6 public at large that autoplay on YouTube contributes to excessive and addictive use. 7 In 2021, YouTube turned autoplay off by default for "all users 13-17." Ex. 1064 at 0252. It also 8 added an option to toggle autoplay on or off on YouTube Kids, setting the default off. Id. This was an 9 important measure. But it came with some serious limitations. First, the default setting only applies to 10 users who are logged-in; and, as discussed below, YouTube is easily and often accessed in a logged-out 11 state. Second, "logged-in" teenagers will only have autoplay defaulted to off if they correctly identified 12 themselves as teenagers upon sign-up. YouTube knows this does not happen for millions of teenage users 13 and (again, as discussed below) it intentionally delayed implementing age inference models to gain a 14 robust understanding of user age. Ex. 1028 at 41 (YouTube estimates there are 69M teens on YouTube); 15 Ex. 1137 at 78 (YouTube asserts that actual teens on the platform are close to of daily average viewers 16 rather than the 4% who declare); Ex. 719 at 3742 (only a small fraction of those who are actually under 17 18 are accurately declaring their age); Ex. 932 at 0478 (2022 email chain discussing and 18 noting that "there are still underage accounts left in the corpus at precision"). Third, YouTube 19 employed dark patterns to encourage users to turn autoplay back on, designing the toggle to be 20 prominently displayed and easy to access. See Ex. 1010 at 288:25-289:11, id. at 292:2-12 (autoplay toggle 21 was "very prominent" and "easy-to-understand"; it was "highly visible…right on the player in mobile" 22 and "right below the player" in desktop). By contrast, and as discussed in later sections, see infra 23 § III.A.2.d.h, YouTube's screen time management tools are buried in a settings page. 24 (b) Infinite Scroll 25 YouTube employs endless scroll features on its platform, including on homepage, Watch Next, 26 and Shorts. See Ex. 792 at 5068; Ex. 793 at 6563; Ex. 1010 at 230:25-231:9, id. at 233:16-235:1; see also 27 Ex. 987 ¶ 203. For example, YouTube Shorts is designed to display in a "Shorts Player," an infinite feed 28 of content modeled after TikTok and Instagram Reels. See Ex. 1172 at 1649, 1665 (internal TikTok 166 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 179 of 235 1 document from 2020 describing "Short-video war" and noting that "users can swipe through YouTube 2 Shorts vertically—just like TikTok."). YouTube is aware that endless scroll features contribute to 3 compulsive and addictive use. See Ex. 715 at 351:9-23 (moving Watch Next feed to infinite scroll 4 increased user time on platform); Ex. 796 at 8133 (experts highlighted that endless scrolling can be 5 detrimental); Ex. 732 at 9252 (concerns about prolonged unintentional use "are loudest on short-form 6 content (more popular with teens) due to its lack of depth and infinite feed experience"). 7 This is echoed by the testimony from Plaintiffs' experts. See Ex. 987 ¶¶ 59-61 (identifying features 8 like infinite scroll as a type of dark pattern designed to maximize engagement and attention); Ex. 1004 ¶ 9 445 ("YouTube's design features, such as likes, comments, notifications, autoplay, and endless scrolling, 10 encourage addictive screen time."); Ex. 983 ¶ 5.a ("continuous scroll takes advantage of multiple aspects 11 of brain development, less developed self-regulation, adolescents' focus on social connection and peer 12 feedback due to social development, and adolescent egocentrism"); Ex. 982 ¶ 194 ("Endless Scroll feature 13 has been linked to prolonging engagement at the cost of displacing children's developmental 14 opportunities"). 15 Finally, as discussed above, the inventor of the infinite scroll, Aza Raskin, has publicly stated that 16 the goal of the feature was to "deliberately keep [users] online for as long as possible." Ex. 252 at 2. At 17 his deposition, he described the infinite scroll as "an intentional removal of stopping cues so that your 18 brain doesn't wake up to catch up with impulses….it creates that hypnologic state where you just keep 19 scrolling." Ex. 1070 at 41:16-21; see id. at 55:21-56:5 (reviewing example of infinite scroll on YouTube). 20 He testified that, "as the designer of infinite scroll, I know that if you do not give the user's brain a chance 21 to catch up with … their impulse, then you can predictively get them to continue to scroll and look at 22 content." Id. at 57:18-58:6; see id. at 55:21-56:5 (reviewing example of infinite scroll on YouTube). "If 23 you had a glass of wine that just kept refilling, like, your mind would never have that moment where you 24 say, oh, I reached the bottom, do I really want another or are you just going to keep sipping and keep 25 sipping." Id. at 24:22-25:1. Sites like YouTube "are all taking great care to preload the content so that the 26 infinite scroll is as smooth as possible." Id. Reflecting on how his invention has been deployed by social 27 media platforms, Raskin stated: "when you get on to a bus or a train or you go into a classroom …and you 28 watch their thumb flicking and flicking and flicking…because of…an invention that I had made that is 167 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 180 of 235 1 now being used for something else, it's very hard. Like that is a deep, hard, sick place in the pit of my 2 stomach when I really get in touch with it." Id. at 65:6-18. 3 YouTube has failed to present any evidence in its summary judgment brief that, at any point, it 4 informed its users, their caretakers, Plaintiffs, or the public at large that endless scroll on YouTube 5 contributes to excessive and addictive use. 6 (c) Algorithm 7 The YouTube recommendation algorithm was launched in 2008 and determines what content will 8 be recommended to users on YouTube. Ex. 1045 at 9. It is designed to keep people on the platform by 9 presenting them with content that it infers (based on an analysis of behavioral signals) will keep them 10 using. Ex. 546 at 4852 ("The user behavior patterns that are of our interest here are the ones we can shape 11 through changes in the recommendation algorithms." (emphasis in original)). Ex. 1045 at 11 12 (recommendations are made by leveraging user-specific and video-specific signals, including the user's 13 watch history, search history, demographics, and activity level); Ex. 776 at 6773 ("There is a lot more we 14 can infer from user's interactions—e.g. what items they scroll past without clicking—and update results 15 in real time."). 16 For many years, YouTube explicitly optimized its recommendation systems to maximize Watch 17 Time above all else, even when doing so meant "essentially ignoring the initial intention the user had 18 when coming to YouTube." Ex. 767 at 5100; see also Ex. 1045 at 9-13; Ex. 767 at 5098-5100; Ex. 711 at 19 3; Ex. 985 ¶ 125; Ex. 767 at 5100 ("All other things being equal, our goal is to increase (video) watch 20 time."). To satisfy YouTube's hunger for its users' time, the algorithm must be constantly updated— 21 hundreds of small changes that add up to large effects on users. YouTube uses an experimentation 22 platform called to measure the effectiveness of these changes. For instance, in 2018, there were 23 changes made to the recommendation system, resulting in a increase in "YouTube Time" (the 24 largest component of which is Watch Time). Ex. 539 at 46. YouTube's users are also its lab experiments. 25 Since 2016, YouTube has included user "satisfaction" alongside watch time as a metric that its 26 recommendation system optimizes for. Ex. 985 ¶ 132. However, as the Districts' algorithm expert makes 27 clear, YouTube measures "satisfaction" in ways that often bottom out in engagement metrics like time 28 spent. See id. ¶¶ 132-137. Similarly, certain YouTube witnesses testified about the company's attempt to 168 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 181 of 235 1 "drive value" for users, including through its recommendation system. Ex. 1003 at 41:2-10, 326:08-14, 2 334:10-23. However, YouTube's employees understood that "the stated goal of value was viewer 3 addiction." Ex. 1116 at 0356 (noting this in section titled "Value prop"); Ex. 1003 at 195:24-196:03. 4 In short, regardless of the words used to describe it, the goal of YouTube's recommendation 5 algorithm has been and continues to be keeping users engaged, i.e. watching videos. See, e.g., Ex. 1045 6 9, 13-14; Ex. 772 at 7889 ("engagement" is still measured primarily by Watch Time or YouTube Time 7 and focuses on "driving greater consumption of videos"); Ex. 982 ¶ 189 ("algorithms are optimized to 8 maximize time on the platform rather than healthy interactions with a person's social network"). 9 Recommendations surfaced by YouTube's algorithm are placed in areas that will draw the most 10 user attention—Home, Watch Next, Search, the Shorts player, and YT Kids Home. Ex. 985 ¶ 123. They 11 are displayed via autoplay and endless scroll features that drive over-consumption. See supra; see, e.g., 12 Ex. 775 at 3; Ex. 788 at 15, 33 (Watch Next panel "encourage[s] viewers to watch more videos" and "has 13 been referred to as a 'rabbit hole'"); see also Ex. 983 ¶ 5.d. ("Recommender algorithms take advantage of 14 multiple aspects of brain development as well as adolescents' less developed self-regulation"). 15 YouTube failed to assess its recommendation system for risks to teens. It was not until 2021 16 (thirteen years after launching the system), that YouTube consulted with experts to identify potential 17 mental health risks to teens. In doing so, YouTube learned that its algorithm can negatively impact teen 18 wellbeing by aggregating and repeating certain types of videos, any one of which could be harmless on 19 an individual basis but harmful in volume. Ex. 1026 at 2617-2620 (reiterating that the problem is watching 20 volume of videos that repeat and reinforce skewed body norms; "too much of this content may lead teens 21 to normalize this and feel pressured to do the same thing because their perception of reality is distorted"). 22 Critically, the root of the harm stems not from the content displayed but from the YouTube 23 recommendation system itself—which, like the recommender systems of YouTube's co-defendants, has 24 a proclivity towards "strong feedback loops." Ex. 1173 at 31; see Ex. 985 ¶ 183. This means that, "[o]nce 25 you watch a few" of a certain kind of video, "your feed might become concentrated [with] a high volume 26 of content that repeat the same message." Ex. 1026 at 2619-20. YouTube understood, for example, that 27 social comparison and body image videos, when viewed in repetition/volume, can skew teen perception 28 of norms and lead to harmful behaviors. See Ex. 1026 at 2617-2620 ("sometimes the Watch Next feed 169 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 182 of 235 1 may have many more videos that repeatedly make negative social comparisons of physical features."); 2 Ex. 1025 at 2611; Ex. 1028 at 2014, 2017; see also Ex. 1023 at 95:21-25 (former senior product manager 3 for health and mental health: "I would say it's a combination of the user kind of urged to watch or choice 4 [sic] to watch more content this way. And then the loop of it feeding that signal into the recommendation 5 system"). In other words, the algorithm was tuned to do whatever it took to keep children's eyes on 6 YouTube, without any regard to what was being shown. But the inevitable outcome of this behavior- 7 maximization was children becoming trapped in echo chambers of ever escalating severity. 8 For example, YouTube understood that a person watching a video presenting a high degree of 9 social comparison would have a 45% chance of receiving Watch Next recommendations for two similar 10 videos. Ex. 1025 at 15. Experts advised that having just three or more such videos recommended in a 11 single session is "too many." Id. at 16. Internally, YouTube acknowledged that "[t]hese concerns have 12 also been escalated in the news by parents and [key opinion leaders]." Ex. 1026 at 2625. 13 YouTube's recognition of the harms caused by its algorithm is further evidenced by Project VIBE 14 ("Volume Impacts Well Being"), whose goal was to ameliorate the harms caused by this algorithmic 15 clustering. Presented on August 10, 2022 by the Youth, Health, and Trust and Safety teams, the initiative 16 had two parts: first, create a classifier to identify categories of content recognized as harming the wellbeing 17 of teens when viewed at high, repetitive volumes; and second, use that classifier to disaggregate harmful 18 algorithmic recommendations and limit its overwhelming rabbit hole effects. Ex. 1028 at 3, 18, 64, 68, 19 76, 117-118; Ex. 1025 at slide 19; Ex. 1001 at 74:4-I10; Ex. 702 at 45:6-50:25. Discussing the effects of 20 VIBE, Dr. Jessica DiVento Dzuban, formerly YouTube's Global Head of Mental Health, offered the 21 "[h]ypothesis …that exposure without VIBE would increase risk. But doing A/B testing on this is likely 22 unethical (as we wouldn't expose someone to risk increasing content if that's our hypothesis)." Ex. 1024 23 at 25:10-13; Ex. 1029 at 212:22-213:2. Stripped of research jargon, this is a YouTube executive 24 hypothesizing that the experience of YouTube without Project VIBE—i.e. the experience that children 25 had received for over a decade—put children's wellbeing at so much risk that that it would be unethical 26 to continue exposing children to it in order to formally test the hypothesis. Indeed, there is no evidence 27 that A/B testing was ever used to measure the efficacy of VIBE. 28 170 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 183 of 235 1 While the existence of Project VIBE demonstrates that YouTube was aware of the risks its 2 algorithms posed to young users, its actual implementation is evidence of YouTube's broader failure to 3 take these risks seriously. When it launched, Project VIBE was initially limited to the Watch Next panels. 4 It did not launch on Shorts—YouTube's most popular surface for teens—until Q4 of 2023. See Ex. 702 5 at 45:6-50:25; Ex. 1029 at 234:1-24. Even then, it only applied to Declared Teens who were also logged 6 into their account at time of use. Id.; see also Ex. 545 at 16 (stating "[d]eclared Teen has low recall. Only 7 of teens on [YouTube] declare 13-17" and "[d]eclared teens only ~3% of [Daily Active Users]"). 8 YouTube did not inform its users, their caretakers, Plaintiffs, or the public at large that its 9 recommendation system was aggregating content in harmful ways and at harmful volumes—nor that its 10 attempted solutions were limited in scope and did not extend to Shorts. See Ex. 1023 at 156:10-19. 11 (d) Age Verification 12 The Children's Online Privacy Protection Act (COPPA) requires platforms like YouTube to, 13 among other provisions, obtain verifiable parental consent if they allow users under the age of 13 on their 14 site. 15 U.S.C. § 6502(b)(1)(A)(ii). If they allow such users, they must also provide parents with the 15 opportunity to review personal information collected from their children and the opportunity to refuse 16 further use or maintenance of such information. 15 U.S.C. § 6502(b)(1)(B). These are in addition to a 17 bevy of other requirements. See, e.g., 16 C.F.R. § 312.4 (requiring posted privacy policies); 16 C.F.R. 18 § 312.5 (mandating mechanisms of parental consent); 16 C.F.R. § 312.6 (establishing parents' right to 19 review information provided by a child); 16 C.F.R. § 312.8 (imposing data retention and deletion 20 requirements). In its summary judgment briefing, YouTube presents no evidence that it complies with 21 these requirements. If anything, YouTube appears to rely solely on the fiction that it does not allow 22 children 13 or younger on its platform. See Exs. 1047, 1067, 1118, 1119, 1120, 1121 (YouTube Terms of 23 Service from 2018 through 2023); Ex. 745 at 32. 24 In reality, a key element of YouTube's mission to embed itself in the daily lives of young users is 25 the ease with which children can access YouTube. Anyone—regardless of age—can access YouTube. No 26 account, screening, age information, or guardrail of any kind is deployed by YouTube to prevent general 27 access. And YouTube is well aware that this unfettered access leads to large numbers of children under 28 the age of 13 (u13) using the platform. See Ex. 724 at 5688 ("vast majority of u13 usage is…happening 171 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 184 of 235 1 anonymously on [YouTube]," e.g., without an account); Ex. 766 at 491:23-492:1 ("YouTube signed out 2 is very easy to access, and it's very likely that some of those users are under 13"). YouTube's failure to 3 effectively prevent children under the age of 13 from accessing the platform, at any time of day, has 4 compounded harm to schools and school districts. Ex. 1022 at 86:19-22, 251:4-20 (noting YouTube was 5 aware it was a source of distraction for students). 6 YouTube makes certain features such as commenting, liking, and subscribing available only to 7 users who are logged in. Ex. 716 at 59:21-60:16 (describing pop-out screen prompting sign-in). But 8 YouTube entices kids to do precisely that, by actively prompting children to create an account and failing 9 to implement any meaningful age gate during the account creation process. See Ex. 987 ¶¶ 42-62; Ex. 715 10 at 324:3-21 (testifying that the logged-in experience had to be more compelling to discourage logged-out 11 usage). Prior to 2016, there was no age-barrier to account creation. Ex. 717 at 1, 6, 8. Since 2016, YouTube 12 has required that users input a date of birth of 13 or older (the "Declared Age") to create an account. Id.; 13 Ex. 718 at 7653. But there is no age validation measure to confirm the birthdate provided, meaning 14 children can simply input an inaccurate birthdate and immediately unlock the full YouTube experience. 15 YouTube is well aware that Declared Age is unreliable. See Ex. 719 at 0742 ("only a small fraction 16 of those who are actually u18 are declaring accurately"); Ex. 720 at 3366 ("Declared isn't a reliable signal 17 to build teen specific models); Ex. 722 at 8465 ("We've heard legitimate concerns that kids under 13 are 18 using the main YouTube site unsupervised….We know it is our responsibility to address how our 19 products are being used" (emphasis in original)). But YouTube failed to implement age validation, 20 knowing it would conflict with its goal of obtaining and retaining young users. See, e.g., Ex. 721 at 21 5186.ECM (acknowledging that "collecting a date of birth (DOB) at sign up for an account is no longer 22 sufficient for products that are deemed not suitable for children" and "Google has to come up with a 23 comprehensive strategy to tackle this problem while ensuring we do not negatively impact the user 24 experience or our products") (emphasis added). 25 YouTube's age verification failures cannot be ascribed to technical capacity, and instead reflect 26 the systemic, years-long deprioritization of safety, even in the face of internal alarm. Since at least 2018, 27 Google's advertising teams have been able to infer age for advertising purposes. See Ex. 1042 at 48:7-14, 28 59:18-20; Ex. 909 at 34:6-35:8. This is because advertisers want to understand who they can target, and 172 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 185 of 235 1 certain products are more suited towards certain people, certain genders, certain ages. See, e.g., Ex. 1010 2 at 50:19-51:10 (interest-based marketing is where "we build an understanding of what interests a user 3 might have based on…where we have visibility into them across the web, and we create interest categories 4 that we associate with those users"); Ex. 909 at 35:17-36:9 (inferred age in Google Ads available "for 5 quite a while now"); id. 64:25-65:3, 66:8-16; Ex. 781 at 2448 (document referencing data column 6 which is age inference). Yet prior to 2019, YouTube's only method of detecting and 7 removing underage users was to manually remove them when employees, through pure happenstance, 8 became aware that a user was under 13 in the course of carrying out other work. See Ex. 1032 at 2. 9 Automated processes could be used to sell things to users, but not to keep them safe. 10 In 2019 YouTube finally created a machine learning classifier to detect u13 accounts, which it 11 called . See id. 2. In theory, once flagged an account as u13, the account would be "auto- 12 actioned." But YouTube placed a number of restrictions on this process, significantly undercutting its 13 efficacy. First, even if accounts are flagged by , the "auto-action" is not deletion. Instead, YouTube 14 sends the users an automatic notice informing them they have 14 days to confirm they are either under 13 15 years of age (in which case, they must set up a supervised account) or over 13 years of age (by submitting 16 proof of, for example, a government ID), or delete the account. See Ex. 780 at 2810; Ex. 718 at 4. Second, 17 review was limited in scope. 18 From November 2019 to November 2020, the u13 19 classifier , and from Feb. 2021 to Jul. 2021 the classifier 20 . See Ex. 780 at 2809. This limitation meant that 21 22 Third, YouTube did not include in the scope of its classifier's 23 review, despite knowing that a Id. at 2813 (2023 24 document discussing need to include in u13 classifier review). Fourth, 25 See Ex. 1122 at 87:19-23; Ex. 1123 at 1. From November 2019 26 to July 2021, flagged accounts would only be 27 . Ex. 1032 at 4; see also, e.g. Ex. 780 at 2810; Ex. 718 at 4. In July 2021, that threshold 28 was . Ex. 1032 at 4. Anything lower— 173 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 186 of 235 1 2 YouTube also placed a 3 Ex. 780 at 2812; see also Ex. 1032 at 4. 4 YouTube knew that these , with internal reports warning 5 that Ex. 780 at 2812 (emphasis 6 added). For example, in 2024, YouTube found that "underage accounts that are actioned by are 7 generally 8 Ex. 780 at 2812 (emphasis added). 9 YouTube further recognized that 10 Ex. 11 780 at 2813. The therefore reflects critical flaws in u13 review: young users were allowed 12 to remain on the platform 13 and thus did not benefit from any age-verification measures 14 to correct their age. But rather than invest resources in Google—one of the 15 wealthiest and most successful technology companies in the world—left the in place. See Ex. 1032 16 at 5. YouTube also never informed parents, teachers, or schools of the fact that its age-verification process 17 was intentionally designed to be attenuated and inefficient. 18 YouTube has the capacity (if not the will) for effective age verification. In 2020, the EU amended 19 its Audio Visual Media Services Directive (AVMSD), requiring YouTube to "establish and operate age 20 verification systems to protect minors or under 18 users from content which may impair the physical, 21 mental or moral development." Ex. 782 at 8. This rendered inadequate YouTube's approach of allowing 22 a user to input a Declared Age to access mature content. See id. at 11-12. So YouTube began developing 23 an age inference model to detect under 18 year old users, which it launched in the EU in late 2020. See 24 Ex. 782 at 12. Among other things, this system leveraged Google demographic signals. See id. at 12; Ex. 25 717 at 1. 26 In September 2021, the United Kingdom's Age Appropriate Design Code (AADC) came into 27 effect, after a twelve month transition period. Ex. 1141 at 16. The AADC required YouTube to "provide 28 either a child-appropriate experience for all users, or apply an 'age assurance' model with a sliding scale 174 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 187 of 235 1 of different levels of age verification depending on nature and risks associated with processing." Ex. 1142 2 at 14. Faced with a requirement in the UK to "verify the age of all users or apply requirements to all age 3 unknown users," Ex. 783 at 8, YouTube chose the former option, and deployed an age inference model to 4 detect users under 18. See Ex. 717 at 1. For users in the UK, as well as the European Economic Area and 5 Switzerland, YouTube began to use inferred age to apply age-appropriate design features discussed 6 elsewhere in this section (e.g. take a break reminder, bedtime reminder, the VIBE classifier, turning 7 autoplay off by default). See Ex. 783 at 28-29; see also Ex. 880 at 9452 (geographic area for "Minor 8 Mode"). This age assurance model did not require any new signals and data collection points. See Ex. 717 9 at 2; Ex. 716 at 32:23-33:2. 10 Meanwhile, back in the United States—where YouTube remained free and clear from such 11 regulatory requirements—it deployed none of these already available age inference measures to protect 12 American teenagers and children. Instead, YouTube continued to rely solely on Declared Age, which it 13 knew to be unreliable. Ex. 719 at 3742. YouTube's decision to do nothing more for American users is 14 egregious and can only be explained by the company's determination to prioritize revenue over user safety. 15 This only began to change in February 2025—well after the initiation of this lawsuit. Ex. 1012 at 16 192:20-194:21 (testimony from CEO Neal Mohan that YouTube was, at this point, "willing to take that 17 financial hit" because "the models have gotten to the point where we could have a pretty good inference 18 model."). On February 11, 2025, YouTube announced that it planned to "use machine learning in 2025 to 19 help us estimate a user's age – distinguishing between younger viewers and adults – to help provide the 20 best and most age appropriate experiences and protections." See Ex. 717 at 18. YouTube's incredibly 21 belated announcement that it would apply age inference to U.S. users, see Ex. 716 at 53:9-54, came five 22 years after it had already launched age inference models in the EU. And it came after years of internal 23 discussion concerning the need for a global model. See Ex. 1044 at 186.ECM; Ex. 784 at 6550; Ex. 719 24 at 3742; Ex. 785 at 3214; Ex. 786 at 18. 25 YouTube has tried to excuse its delay on the grounds that age inference models are "technically 26 challenging." Ex. 766 at 553:14-554:4. But, as discussed above, its parent company Google already had 27 the data points needed to infer age, having done so for years to make additional profit in the advertising 28 space. And YouTube had already launched age inference models in other jurisdictions for the express 175 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 188 of 235 1 purpose of inferring users under the age of 18. Technical feasibility is simply not an excuse. 2 YouTube has also maintained that, because age inference models are probabilistic, they needed to 3 be tested and configured to ensure an appropriate level of accuracy. For example, YouTube's CEO 4 expressed concerns that incorrectly inferring an adult user as a teen would lead to a degraded experience 5 for the adult because, for example, they might not receive personalized advertising. See Ex. 1012 at 6 199:11-15 (attributing delay in part to concern that "some number of users who might be above 18 are 7 now for some reason classified as below 18. And so, therefore, they get a degraded [non-personalized] ad 8 experience"); see also Ex. 766 at 677:13-679:17 (concern from senior director that false positives would 9 be disruptive and annoying for adults) . 10 But this parade of horribles is a red herring and, even if true, a self-constructed barrier to avoid 11 implementing age inference in the United States. YouTube can (and has, in response to the AADC and 12 AVMSD) applied the output of its age inference models for the limited purpose of turning on protective 13 settings for young people—which is beneficial, not harmful, to the overall user experience. Moreover, 14 inherent in this excuse is the implication that perfection should supersede protection. In YouTube's world, 15 the risk that it could potentially over-designate adults as minors—and apply protective features to them 16 that would moderate usage—is simply too high a cost. This reasoning is in direct contradiction with what 17 YouTube itself recognizes is its obligation to protect young users from harms on its platform. 18 The real reason for YouTube's delay is more straightforward: YouTube feared a loss of revenue. 19 In the run-up to YouTube's February 2025 announcement, the company forecasted the potential revenue 20 implications of rolling out age inference globally. It determined the cost could range from 21 depending on the amount of "readiness work" that preceded the change. Ex. 787 at 8_022; 22 Ex. 1012 at 206:5-18. 23 (e) Barriers to Exit 24 Even as YouTube made access to its platform incredibly easy—including through a logged-out 25 state that requires no age verification—it has made account deletion incredibly difficult. See Ex. 987 ¶¶ 26 219-222 (YouTube account deletion process exemplifies "high-level dark patterns, which impede users 27 from exercising autonomy"). In fact, the process for deleting a YouTube account is so difficult that in one 28 "Google Account Deletion Flow" video produced by YouTube in this litigation, a system engineer was 176 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 189 of 235 1 unable to successfully delete an account—instead creating, and then deleting, an additional Google+ 2 account. See Ex. 723 (video). The marked difference between ease of access, and difficulty of deletion, 3 demonstrates that YouTube can and does leverage design barriers to serve YouTube's bottom line. 4 (f) Parental Controls 5 As detailed in this section, YouTube's purported parental controls and supervision tools were, and 6 remain, woefully deficient in terms of addressing addictive and compulsive use. 7 Family Link (launched to the public in 2017, see Ex. 1143 at 1942) is not a YouTube specific 8 parental control/supervision feature. See Ex. 730 at 3; Ex. 766 at 472:8-22. It is a complicated set of 9 controls that is limited to Google accounts only; that is buried in the Google accounts setting page; and 10 that varies across user devices (Android vs. iOS) and platforms (mobile versus desktop). See Ex. 1015 at 11 1210; Ex. 730 at 3. This makes Family Link ineffective as a screentime monitoring or supervision tool for 12 YouTube, a fact YouTube acknowledges. See Ex. 1016 at 4601 ("Google and YT's current offerings are 13 inadequate"); Ex. 1015 at 1210 (depending on device, children "can only be partially supervised"). 14 YouTube knew that usage of Family Link was low. Ex. 1017 at 9 ("Family Link among parents is low 15 . It knew that variations across devices meant gaps in service. Ex. 1016 at 4601 (Family Link 16 controls do not apply to the of U.S. users on iOS). And it knew that options buried in account settings 17 were difficult to access. Id. Furthermore, YouTube understood that children could easily "circumvent 18 policy restrictions and parental controls" of Family Link by simply accessing YouTube in a logged-out 19 state (i.e. not signed into an account). Ex. 1018 at 0294.ECM. This is assuming teens even opt in to Family 20 Link in the first place—as the setting requires their agreement. See Ex. 1019 at 126:8-25. 21 Importantly, Family Link did not provide parents with an adequate way to monitor and control 22 their children's usage of YouTube; as such it does nothing to help parents address addictive and 23 compulsive use. See Ex. 1016 at 4601. 24 Supervised Experience on YouTube Main (SupeX) was launched in 2021. See Ex. 730 at 3. As 25 discussed above, SupeX was designed to address a perceived product gap for 8 to 12 year olds aging out 26 of YouTube Kids and seeking an older experience. See, e.g., Ex. 731 at 6489 ("SupeX has the content 27 missing" from YouTube Kids). SupeX was and is ineffective as a supervision tool. As with Family Link, 28 it is easily circumventable: Anyone can access YouTube without an account, and tweens can easily create 177 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 190 of 235 1 an account without parental supervision. See supra § III.A.2.d.5.d. 2 Moreover, the supervision aspects of SupeX are woefully deficient in terms of addressing addictive 3 and compulsive use. SupeX controls are focused on content control—e.g., choosing a specific content 4 "corpus," channel blocking, and disabling or deleting a child's search and watch history. See Ex. 730 at 5 3-4. Those are all content-focused restrictions; they do not assist a parent in navigating the content- 6 agnostic problem of addiction. And while SupeX does allow parents to turn off autoplay, it retains many 7 of the other addictive features described in this section, such as an algorithmic recommendation system, 8 notifications, filters, and an infinite scroll in Shorts, Watch Next, and elsewhere. Moreover, while 9 YouTube knew that parents wanted some ability to control their children's screen time, they did not 10 include any screen time limit within SupeX. See Ex. 1007 at 6134-35; Ex. 1008 at 5271 (2022 report 11 indicating that parents thought a "Supervised channel" was one that let them "limit screen-time use"). 12 Parent Supervised Experience for Teens was not launched until 2024. See Ex. 730 at 4-5. It is 13 ineffective as a supervision tool. First, it is an opt-in experience for the teen; they can stop it at any time. 14 See Ex. 1020 at 9045; Ex. 1021 at 211. Second, even if opted-in for one account, teens can circumvent the 15 supervision by just creating other accounts. Third, with respect to addressing addictive and compulsive 16 use, Parent Supervised Experience for Teens does nothing. It has limited tools only for "teen users who 17 post content." Ex. 730 at 4. In those cases, parents will be able to see channel activity and receive a 18 notification when the teen posts public video or starts a livestream. See id.; Ex. 1019 at 31:20-32:1. There 19 are no controls as to screen time or any visibility for parents into their teen's watch time history. See Ex. 20 1031 at 4; Id. at 28:4-30:14, 87:12-25. 21 Given that YouTube's parental controls don't actually give parents control, it's fair to ask what 22 they're actually for. The answer? To fulfill YouTube's business objectives. As noted above, SupeX was 23 designed principally to help YouTube capture the market of "tweens" too old for YouTube Kids. See 24 supra § III.A.2.d.2. Likewise, Family Link helped Google acquire young users by creating a path for them 25 to create a Google Account (otherwise not permitted for those under 13). See Ex. 1139 at 9-10, 25; Ex. 26 765 at 42 (Deep Dive with Neal and Matthew) (noting "parents can use Family Link to create accounts 27 for kids under 13"). While presented under the guise of "protection," supervision features like these were 28 in fact just tools for YouTube to profit more off kids. 178 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 191 of 235 1 Finally, a critical component of effective control and supervision is to inform the decisionmaker 2 of the impacts of their choice. YouTube never warned parents or kids about the addictive designs on its 3 platform, and it never warned parents of the alarming gaps in the controls and supervision tools presented. 4 See infra § III.A.2.d.7. Individuals cannot judge the level of vigilance and action needed if they are not 5 informed of the scope of harm and if they are misled about the efficacy of the protections offered. 6 (g) Notifications 7 Notifications are a "critical" part of YouTube and an important way it drives engagement. Ex. 775 8 at 14; Ex. 798 at 0466 (describing use of "reengagement emails targeted at less engaged users."). 9 YouTube's commitment to reengaging users through notifications is tremendous. Ex. 1051 at 216:21- 10 218:21 (increase in notifications a core driver and "unlock" for engagement growth and business value). 11 In 2018, its Growth Team listed as a top objective growing notifications' reach and impact, setting a goal 12 of increasing daily notification openers to and daily active viewers from algorithmic 13 notifications to . See Ex. 1048 at Growth Team Tab, Row 4. By 2021, YouTube was sending 14 notifications daily—up from notifications per day sent in 2016. See Ex. 1049 at 15 9. This amounted to algorithmically-generated, growth-oriented notifications. Id. at 9. 16 YouTube is aware that notifications contribute to excessive and addictive use, as they prey on 17 users' "fear of missing out." See Ex. 775 at 6 (wellness factors diagram showing a connection between 18 addiction, notifications and FOMO); id. at 10, 14 (receiving notifications causes users "to feel like they 19 must be aware of what is happening on the platform" and tempts users to watch videos the moment they 20 are uploaded). Self-reports from female users confirmed that morning and evening app use was triggered 21 or driven by notifications, especially for teens, and that, relatedly, bedtime triggered notification checks 22 commonly resulted in "mindless scrolling…before falling asleep." Ex. 1050 at 4675-76; see also Ex. 799 23 at 5968 ("Notifications can be distracting and overwhelming, especially close to bedtime"). Notably, 24 despite YouTube researchers finding that "notifications... contribute to addiction," YouTube sent 25 notifications to students during instructional school time. Ex. 775 at 10, 14. YouTube also never warned 26 the public of the risk that its notifications posed to young users' wellbeing and instruction. Ex. 997 ¶¶ 155, 27 159-160, 166-67, 186-187, 191-92, 195-196, 201, 202. 28 179 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 192 of 235 1 The Districts' experts agree that YouTube's notifications present significant risk to teenage users 2 in particular. See Ex. 1004 ¶ 52 ("Notifications and alerts that prompt repeated engagement leverage the 3 variable timing of cues and rewards to reinforce compulsive checking behaviors"); Ex. 987 ¶ 53 4 (notifications are a type of dark pattern and "attention capture damaging pattern" that cause reactivation 5 of engagement); Ex. 983 ¶¶ 5.c ("Push notifications take advantage of multiple aspects of adolescent brain 6 development, less developed self-regulation, social development, and adolescent egocentrism"). 7 (h) Screen Time Management 8 In 2018, YouTube launched a set of three tools to address digital wellbeing: (1) Take a Break 9 Reminder, (2) Time Watched Profile, and (3) Notification Clusters. See Ex. 1033 at 0058. In 2020, it 10 launched another similar tool, (4) Bedtime Reminders. See Ex. 730 at 5; Ex. 1034 at 9501. All four are 11 opt-in tools, meaning that to access the tools, users would first have to know of their existence, then 12 navigate to a separate setting screen, find the setting, and opt-in to receive the treatment. This high user 13 friction ensured that few would opt-in, deeply limiting the tools reach and efficacy. Cf. Ex. 989 ¶ 34; see 14 also Ex. 1022 at 199:7-201:25; see generally Ex. 1036; Ex. 987 ¶ 205. 15 Take a Break Reminder. Take a Break Reminder was aimed at heavy habitual use. See Ex. 1034 16 at 9499. At launch, it was an opt-in setting that allowed users to set a reminder to take a break after n 17 minutes of watching videos. See Ex. 799 at 5966; Ex. 1034 at 9499. In addition to being hard to find, even 18 if enabled it was easy to dismiss. Indeed, "the reminders are presented in way that incentivizes dismissal 19 and continued content consumption, rather than choosing to end a session." Ex. 987 ¶ 206 (emphasis 20 added). That is so because, at the time of launch, the reminder literally highlighted "dismiss" as an 21 option—"over the user exiting the platform." Ex. 987 ¶ 206. This design employs the dark patterns known 22 as "interface interference" and "visual prominence," id., encouraging users to dismiss their reminders and 23 continue using YouTube. See Ex. 1033 at 0060. 24 Time Watched Profile. Time Watched Profile was aimed at heavy habitual use. See Ex. 1034 at 25 9500. It is also an opt-in tool, meaning it is subject to the same friction described above. Even if accessed, 26 it is a passive tool that simply allows users to check their average watch time. See id. The tool provides 27 no interruption, suggestion, prompt, or guidance regarding continued use of the platform. 28 180 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 193 of 235 1 Notification Clusters. Notification Clusters allowed users to modify when they received 2 notifications. See Ex. 1033 at 060. At launch, it was an opt-in setting for users to bundle push notifications 3 into a single daily notification and set a specific time to receive a notification digest. See Ex. 799 at 5966. 4 Again, it was a hard-to-find opt-in tool, ensuring few would find out about it and turn it on. See Ex. 1033 5 at 0060. In June 2018, YouTube noted that users had enabled the notifications schedule digest. See 6 Ex. 692 at 9739. In 2018, YouTube had over a billion active users, meaning that the opt-in rate for 7 notifications clusters was just . This is not surprising given YouTube's "success criteria"—ensuring 8 the feature had a "neutral" impact on watch time. Ex. 777 at 6605. 9 Bedtime Reminders. This tool was launched to address late night use. See Ex. 730 at 5; Ex. 1034 10 at 9501. Users who opt-in can set specific times to be reminded to stop watching videos and go to bed. 11 See Ex. 1034 at 9501. As with the other tools described above, Bedtime Reminders was hard to find and, 12 even if enabled, easy to dismiss. The reminder appears with two options, "dismiss" or "snooze," and 13 content continues to be displayed. Ex. 1063 at 17, 22. There is no option to immediately leave the YouTube 14 platform. Further, Bedtime Reminders was only available on certain devices and, up until 2022, not 15 available for Shorts (despite Shorts being YouTube's flagship teen-oriented product). See Ex. 730 at 5. 16 Take a Break Reminders, Time Watched Profile, Notification Clusters, and Bedtime Reminders 17 are also all self-policing tools, allowing users to set their own guardrails (or none at all). They are, 18 therefore, unlikely to have a meaningful impact on young users, who YouTube knows have poor self- 19 regulation and impulse control. Ex. 1037 at 41 (recognizing that teens and tweens are in a unique 20 developmental stage and "Self-regulation is not as developed in their prefrontal cortex, so they are less 21 able to moderate and self-reflect in the moment."). 22 In 2021, YouTube changed the settings for Take a Break and Bedtime Reminders, turning them 23 on by default for "all users 13-17." Ex. 1064 at 0252. This was a modest step in the right direction. 24 However, it did not change the fundamental limitation of these tools—they place the onus on young users 25 to self-police and self-regulate, by actually honoring the reminders they're given. As any parent of a tween 26 or teen knows firsthand, "although more developed than children, adolescents are still developing self- 27 regulation, meaning that they may still struggle to control impulses and inhibit urges." Ex. 983 ¶ 47. 28 YouTube tapped into this aspect of adolescent psychology by encouraging teen users to "dismiss" or 181 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 194 of 235 1 "snooze" the reminders provided. 2 YouTube has been happy to publicly tout these digital wellbeing tools, but it is far less focused on 3 their efficacy. Indeed, YouTube does not track whether users act on breaks or bedtime reminders—one 4 witness testified, "That is unknowable" and "We don't track that." Ex. 1031 at 57:7-13, 59:9-20. (As 5 discussed above, this is not "unknowable" for Meta, which does track adoption rates for its comparable 6 screen time management tools. Those rates confirm that opt-in tools are very rarely used. See supra 7 § III.A.2.a.c.) One employee who inquired about the success of YouTube's digital wellbeing tools was 8 sent links tallying the number of users who saw the reminders—but no meaningful metric of their actual 9 effectiveness in limiting use. Ex. 777 at 6605; see also Ex. 779 at 31; Ex. 715 at 457:25-458:23; Ex. 1022 10 at 156:3-20; 254:3-255:23; Ex. 1031 at 58:12-19, 59:9-17 (did not track if people actually turned off 11 YouTube app after receiving a take a break or bedtime reminder). But YouTube continues to parade these 12 tools before users, their caregivers, and teachers, without warning them that there is no reason to believe 13 they meaningfully help users reduce screen time. 14 Tellingly, YouTube does track the impact of its screen time management tools on its bottom-line 15 business metrics— per Ex. 778 at 11) and . As one 16 document concedes: "We also check when rolling out new features 17 to make sure we didn't break anything. They're expected to be neutral though." Ex. 777 at 6605. The idea 18 that screen time management tools would have a 19 is counterintuitive, to say the least. Ex. 982 ¶ 177-178. And it suggests why YouTube declined 20 to measure the efficacy of these tools—it knew they were useless. 21 (i) Beauty Filters 22 YouTube provides users with augmented reality (AR) effects and filters that users can apply to 23 modify their videos. Ex. 1013 at 73:4-25. This includes filters that can change backgrounds and apply fun 24 masks. It has also included filters that can change a user's appearance—by adjusting eye color, brightening 25 teeth, smoothing skin texture, and changing skin tones. Id. at 73:4-13. YouTube launched these filters 26 because it knows they are particularly attractive to young users. Ex. 1055 at 9400-01 (email discussing 27 TikTok's "Bold glamour" filter attracting "millions"); Ex. 1065 at 6687-88. It also knows these filters 28 "drive video creation and consumption." Ex. 1072 at 6050 (funding Effects team to accelerate this); see 182 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 195 of 235 1 also Ex. 701 at 1277 (2018: "If Snapchat and Instagram are any indication, creation follows new effects."). 2 In 2021, YouTube introduced these appearance-altering effects to Shorts, Ex. 693 at 2908, despite 3 knowing it is primarily used by adolescents, Ex. 844 at 4694; Ex. 858 at 3; Ex. 733 at 8317 ("Shorts is our 4 big thing for teen appeal"). As recently as 2024, YouTube was exploring how it could expand visual 5 effects, including those primarily focused on appearance enhancements. See Ex. 1057 at 1837, 1849. 6 YouTube keeps these effects and filters on its platform despite knowing they can be dangerous to 7 youth mental health. Jessica DiVento Dzuban, formerly YouTube's Global Head of Mental Health (and 8 now TikTok's Global Head of Mental Health Policy) testified that she spoke with numerous youth 9 advocacy groups, including the American Academy of Pediatrics, regarding parents' concerns that 10 YouTube was affecting their child's mental health. Ex. 1024 at 81:9-82:6. Dr. Dzuban specifically recalled 11 the discussion centering on body image and exposure to instructions on how to engage in risky behavior. 12 Id. Despite those conversations, YouTube provides beauty filters for use by minors, Ex. 1013 at 73:15- 13 74:13, and allows filters created by third parties, id. at 75:14-17. 14 YouTube knows that appearance altering filters encourage unrealistic and harmful beauty 15 standards and perpetuate cultural and racial biases. See, e.g., Ex. 1053 at 5593 (encouraging unrealistic 16 standards); Ex. 1065 at 6687 (perpetuating bias); Ex. 1055 at 9400 (advancing unhealthy ideals). Indeed, 17 its own research established the profound psychological dangers of these tools. See Ex. 1073 at 31-32 18 (July 2023 internal presentation identifying specific teen wellbeing concerns pre-dating facial deformation 19 effects and noting, with increasingly sophisticated technology, "it is reasonable to predict negative 20 outcomes in line with or potentially worse than what we've previously seen."); see, e.g., Ex. 1074 at 7 21 ("expert researchers found beauty filters and negative social comparison may be correlated with negative 22 wellbeing, especially in youth" and "the pervasive nature of filtered images regularly trigger body 23 dysmorphia."). 24 One YouTube user research study from July 2023 documented that "participants talked at 25 length about how facial deformation effects" made them feel worse about their natural appearance." Ex. 26 1073 at 33; see id. at 29 (deformation effects are those that can only be replicated in real life through 27 surgical intervention, if at all). Users reported not only individual psychological harm, but "social stigma 28 and shame that accompanies the use of facial deformation effects." Id. at 34. YouTube was clear-eyed 183 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 196 of 235 1 about how these effects could result in social anxiety and isolation, describing the following "stages of 2 impact": "I cannot meet people in real life and disappoint them," "I cannot show my real self anymore," 3 "I need to also look this way," and "everyone's perfect but me." Ex. 1075 at 6913. All of this is aligned 4 with the Districts' experts' testimony about beauty filters. See e.g., Ex. 982 ¶ 195 (because of filters, "users 5 are now exposed to 'manipulated' photos that depict unrealistic (and in some cases impossible) body 6 image standards"); Ex. 983 ¶ 5.f ("Filters and other image manipulation options take advantage of pubertal 7 development, adolescent egocentrism, and social development"); Ex. 1053 at 5593-95. 8 YouTube's decision to allow adolescents to utilize appearance-altering beauty filters, despite this 9 research, is a cynical prioritization of profits over child safety. 10 (6) YouTube's misrepresentations and manipulations 11 On August 10, 2021, YouTube publicly announced the launch of "new safety and digital wellbeing 12 options for younger people on YouTube," including certain of the screen time management tools discussed 13 above. Ex. 872 at 1. In a blog post, YouTube touted its "ongoing efforts to create age-appropriate 14 experiences for young people on YouTube," observing "Young people are our future." Id at 4. 15 Unfortunately, while YouTube has been public and vocal about its safety features (inadequate and 16 lackluster as they are), behind closed doors it has undermined those features by ramping up efforts to 17 increase youth consumption. For example, as the same time YouTube was announcing the launch of 18 Bedtime Reminders, it was also finalizing the design of an infinite scroll experience on mobile Watch 19 Next. Ex. 792 at 5068. At the same time YouTube offered tools to reduce distraction by "cluster[ing] 20 notifications," it was setting a goal to increase the total number of notifications sent. Ex. 1040 at 9; Ex. 21 1048 at Growth Tab, Row 4. YouTube's well-being tools could not and did not solve the fundamental 22 problem that the platform is designed to reduce rational decision making and optimize for more Watch 23 Time. Ex. 1041 at 27. 24 Moreover, YouTube's announcements about keeping safe "users ages 13-17 on YouTube," Ex. 25 872 at 2; Ex. 761 at 2 (PTA editorial); Ex. 702 at 51:8-13; blow past the fact that YouTube has no idea 26 who those users are—and has deliberately avoided finding out. Ex. 1012 at 105:19-24 (only a small 27 fraction of those under 18 accurately declare their ages). YouTube could have done better, as it already 28 was outside the United States. Id. at 143:22-144:23. 184 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 197 of 235 1 (7) YouTube's failure to warn 2 YouTube failed to warn parents, the Districts, and their students about the dangers caused by the 3 YouTube platform to young users. It knew that school administrators and parents wanted more effective 4 time management controls, to minimize distraction for students and enforce responsible use during the 5 school day. Ex. 1062 at 0828-29; Ex. 766 at 462:14-469:22; 470:2-20, 476:21-480:3, 471:3-472:22; Ex. 6 764 at 2982-83. It knew that students were easily able to bypass existing tools like MRM, yet it failed to 7 alert schools, including the Districts, to this vulnerability. See Ex. 740 at 250:6-11; 252:16-256:3, 282:2- 8 283:1. 9 Meanwhile, YouTube knew that its persistent focus on increasing Watch Time was having the 10 predictable and intended effect of keeping teens on its platform late into the night—driving a loss of sleep 11 and, from there, distraction, loss of productivity, anxiety, and depression. Ex. 1003 at 227:6-9, 230:13-19; 12 Ex. 1058 at 2685 (data reporting that approximately 7 percent of teens watch YouTube past midnight on 13 school nights, and up to 30 percent note loss of sleep); Ex. 1059 at 33 (finding most people need 7 to 9 14 hours of sleep to be productive but blue light blocks the brain's signals to sleep); Ex. 738 at 15-17, 35, 40, 15 42; Ex. 774 at 14-25 ("disabling or limiting Autoplay during the night could result in sleep savings"); Ex. 16 775 at 6; Ex. 1058 at 2685 ("~7% of teens on YT watch past midnight on school nights. Teen 'night owls' 17 were 88% more likely to have emotional & behavioral problems....30% of users 18-24 say YouTube has 18 cut into sleep"). And while it knew its heavily advertised "Take a Break" and "Bedtime" reminders were 19 minimally effective, it said nothing. Ex. 1061 at 43; Ex. 732 at 9261. 20 It is reasonably foreseeable that YouTube's failure to warn would harm the Districts, by requiring 21 them to divert limited resources to manage YouTube-created problems. 22 3. Defendants' conduct substantially interferes with the public right to health, safety, and education in Arizona, Georgia, Kentucky, and Maryland 23 In the preceding pages, the Districts have illustrated the scale of Defendants' wrongdoing. 24 Defendants' conduct breaches their duty of care to the Districts and is legally negligent. It also constitutes 25 a significant interference with the public's right to health, safety, and education. The Court has permitted 26 Breathitt, Tucson, DeKalb, and Harford to seek remedies for that interference through a public nuisance 27 cause of action. See In re Soc. Media Adolescent Addiction/Pers. Inj. Prods. Liab. Litig., 2024 WL 28 185 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 198 of 235 1 5694360, at *17 (N.D. Cal. Nov. 15, 2024) ("Public Nuisance MTD Order"). 2 Defendants argue in a cursory fashion that these Districts have not demonstrated a significant 3 interference with a public right. Defs.' Breathitt Mot. at 12-13 ("Breathitt must come forward with 4 evidence that each Defendant's actionable conduct was a direct cause of a significant interference with a 5 public right…"); Defs.' Dekalb Mot. at 12-14 (similar); Defs.' Tucson Mot. at 15-17 (similar); Defs.' 6 Harford Mot. at 15-16 (similar). As an initial matter. Defendants waived this argument by failing to 7 develop it. Beasley v. Astrue, 2011 WL 1327130, at *2 (W.D. Wash. Mar. 24, 2011) (cursorily raising an 8 issue without any supporting analysis or explanation waives the issue raised). Defendants' argument also 9 fails on the merits. Defendants cite no cases granting summary judgment on the issue of whether an 10 interference with public rights is "significant." That is unsurprising, because the significance of a 11 defendant's contribution to a public nuisance is a fact-intensive "question properly left to the jury." In re 12 Nat'l Prescrip. Opiate Litig., 589 F. Supp. 3d 790, 809 (N.D. Ohio 2022). 13 In the preceding pages of this omnibus brief, and in each District's accompanying submission, 14 each District has adduced evidence that the design of Defendants' platforms has led millions of young 15 people to experience compulsive use and attendant mental health issues, which significantly disrupts and 16 harms school operations and the school environment, and frustrates the ability of schools to achieve their 17 mandate of educating students. Each District has also put forth evidence that Defendants' conduct 18 foreseeably caused it to expend and divert resources to address these problems. See, e.g., Ex. 1000 ¶¶ 5, 19 123. This is precisely the type of evidence this Court found may support a significant interference with 20 the public right to health and safety. MTD Order, 754 F. Supp. 3d at 956-59; Public Nuisance Order, 2024 21 WL 5694360, at *14 & n.20. Moreover, while the Court declined to address the public right to education 22 at the pleading stage, the evidence set forth in this submission and the Districts' accompanying 23 submissions shows that Defendants' conduct violated the public right to education in Arizona, Georgia, 24 Kentucky, and Maryland. Public Nuisance MTD Order, 2024 WL 5694360, at *14. All four states 25 recognize a public right to education. Az. Const. art. XI, § 1 et seq.; Ga. Const. art. VIII, § 1 et seq.; Ky. 26 Const. § 183; Md. Const. art. VIII, § 1 et seq. 27 This evidence is sufficient for a jury to determine there has been a "significant interference" in a 28 public right and accordingly, creates a triable issue of fact under Arizona, Georgia, Kentucky, and 186 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 199 of 235 1 Maryland nuisance law. Nolan v. Starlight Pines Homeowners Ass'n, 216 Ariz. 482, 489 (Ct. App. 2007) 2 (significant interference means something "real and appreciable" that is "more than slight inconvenience 3 or petty annoyance") (citing Restatement (Second) of Torts § 821F); see also Blondell v. Courtney Station 4 300 LLC, 865 S.E.2d 589, 601 (Ga. Ct. App. 2021). 5 Given that the Districts have proffered evidence of Defendants' significant interference with public 6 rights to health, safety, and education, Defendants are not entitled to summary adjudication on this issue. 7 B. Defendants' causation arguments are premised on disputed facts and legal errors. 8 1. Defendants improperly raise the evidentiary bar in requiring Plaintiffs to quantify specific student usage 9 10 In each bellwether state, questions of actual and proximate causation are for the jury to decide, 11 unless the evidence supports only one inference, such that no reasonable juror could find for the plaintiff. 12 Torres v. Jai Dining Servs. (Phoenix) Inc., 497 P.3d 481, 486 (Ariz. 2021); Johnson v. Wood, 2025 WL 13 2715888, at *4 (Ga. Ct. App. Sept. 24, 2025); Pathways, Inc. v. Hammons, 113 S.W.3d 85, 89 (Ky. 2003); 14 Pittway Corp. v. Collins, 973 A.2d 771, 253 (Md. Ct. App. 2009); Gilbert v. Stewart, 255 A.3d 1101, 15 1113-14 (N.J. 2021); Wickersham v. Ford Motor Co., 853 S.E.2d 329, 390–91 (S.C. 2020). "Only in plain 16 and undisputed cases should a court even consider resolving such fact-intensive questions as proximate 17 causation on summary judgment." Meraz v. Ford Motor Co., 2014 WL 12558123, at *12 (C.D. Cal. June 18 13, 2014) (cleaned up). Defendants have not come close to meeting this high standard. 19 Defendants argue that causation can only be demonstrated if the Districts can precisely quantify 20 the exact harm caused by each Defendant by furnishing data regarding individual student use of each 21 individual feature of each individual platform. Defs.' Tucson Mot. at 5–7, 14–15; Defs.' Dekalb Mot. at 22 5, 12–13; Defs.' Breathitt Mot. at 12–13; Defs.' Harford Mot. at 14–15; Defs.' Irvington Mot. at 14–15; 23 Defs.' Charleston Mot. at 14–15. This argument fails for several reasons. 24 First, Defendants' extreme position—that schools should be able to produce quantifiable evidence 25 measuring their students' use by platform, duration, and even feature—does not reflect the law, and is not 26 consistent with the facts of this case. It is for the jury to decide whether the Districts' evidence is sufficient 27 to establish causation, not for Defendants to dictate the form of that that evidence. In City & Cnty. of San 28 Francisco v. Purdue Pharma L.P., the court rejected the argument that proof of causation required tracing 187 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 200 of 235 1 each defendant's individual prescription all the way through to the end-user's specific harm, because "the 2 substantial factor test does not require 'the kind of incontrovertible linkage proposed by'" Defendants. 3 620 F. Supp. 3d 936, 1003 (N.D. Cal. 2022). The court found "it is difficult to take seriously the argument 4 that the failure to identify specific" such data "is fatal to Plaintiff's case." Id. at 1004, 1004 n.31. So too 5 here. First, there are legal constraints on schools' ability to obtain student usage data. Moreover, as a 6 practical matter, schools are in the business of teaching and caring for students, not meticulously 7 cataloging each instance of social media usage. As Dr. Hoover has explained, school districts are resource- 8 constrained institutions and "existing [school district data] systems often do not track the types of internal 9 service referrals, behavioral disruptions, or emotional dysregulation commonly tied to social media use." 10 Ex. 1103 ¶ 14; Ex. 1102 ¶14; Ex. 1101 ¶14; Ex. 1104 ¶ 14; Ex. 1105 ¶ 14; Ex. 1100 ¶ 16. In a telling 11 contradiction, Defendants oppose the remedy of better data infrastructure for the Districts to track such 12 events. See Defs.' Tuscon Mot. at 11; Defs.' Deklab Mot. at 8–9; Defs.' Breathitt Mot. at 9; Defs.' Harford 13 Mot. at 10; Defs.' Irvington Mot. at 9; Defs.' Charleston Mot. at 9–10. 14 Second, Defendants' argument ignores the substantial evidence the Districts have produced 15 through fact and expert testimony, showing that each District was foreseeably harmed by widespread, 16 problematic use of Defendants' platforms. See supra § III.A.2; see Ex. 1000 ¶ 30; Ex. 656 ¶¶ 54-75; Ex. 17 982 at ¶¶ 269-530; Ex. 1004 ¶¶ 311-50; Ex. 1005 ¶¶ 5.a-5.j; Ex. 981 ¶¶ 282-367; Ex. 1095 ¶¶ 48-52; Ex. 18 1096 ¶¶ 49-51; Ex. 1093 ¶¶ 48-52; Ex. 1091 ¶¶ 52-54; Ex. 1092 ¶¶ 46-50; Ex. 1094 ¶¶ 51-55; Ex. 656 19 ¶¶ 54-75. This evidence incorporates a broad array of facts that support the schools' claims and is more 20 than sufficient to create disputed issues of material fact. In re Nat'l Prescription Opiate Litig., 2019 WL 21 4178617, at *3 (N.D. Ohio Sept. 3, 2019) (denying summary judgment and rejecting argument that 22 plaintiffs could not prove causation against multiple defendants who all contributed to injuries). 23 Third, Defendants are mistaken in arguing that references to "social media" by administrators, 24 teachers, and counselors are not probative. Defendants are not obscure, niche platforms—they are the 25 most popular social media platforms used by American youth. Ex. 981 ¶¶ 284-87, 290, 329-30; Ex. 1038 26 at 1-2. A reasonable jury could infer that when educators and students refer to "social media," they mean 27 Instagram, TikTok, YouTube, and Snapchat—even if each reference does not name the platforms 28 explicitly. See Ex. 1000 ¶ 30 (Dr. Hoover explaining that "[a]ccording to a recent Pew Research Center 188 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 201 of 235 1 report, over a third of teens say they are using the top five online platforms (YouTube, TikTok, Instagram, 2 Snapchat and Facebook) almost constantly"); Ex. 1097A at 203:7-24 ("[S]ince you've asked about the 3 four defendants' platforms, those are … by the work that I do with students and schools, the ones that are 4 most often brought up and used … When I speak about social media, [in my reports] it's talking about the 5 platforms that young people are using, and those happen to be the defendants' platforms."). Even if these 6 statements were viewed as "circumstantial evidence," that is enough: "California courts have repeatedly 7 held that causation may be reasonably inferred from circumstantial evidence in the context of mass torts." 8 City & Cnty. of San Francisco v. Purdue Pharma L.P., 620 F. Supp. 3d 936, 1002 (N.D. Cal. 2022). The 9 same principle applies under the laws of every bellwether state.8 And at summary judgment, all inferences 10 must be drawn in favor of the Districts. Hunt v. City of L.A., 638 F.3d 703, 709 (9th Cir. 2011). 11 Fourth, Defendants are wrong that the Districts must produce evidence apportioning a precise 12 percentage of liability to each Defendant. Determining the relative significance of each Defendant's 13 conduct is a question for the jury. See In re Nat'l Prescription Opiate Litig., 589 F. Supp. 3d 790, 808 14 (N.D. Ohio 2022) ("Determining the significance of each Defendant's . . . conduct [in causing the 15 nuisance] was a question properly left to the jury); see also Social Media Cases, 2025 WL 2807828, at *7 16 (Cal. Super. Sep. 22, 2025) ("Determining whether the substance or instrumentality upon which liability 17 is premised caused a particular plaintiff's harm is . . . the role of the jury."). A nearly identical argument 18 was rejected in the Opioids MDL. There, a network of manufacturers, distributors, and pharmacies each 19 contributed to a nationwide public nuisance. The defendants argued that the local government plaintiffs 20 could not prove the causal contribution of any individual defendant to any particular county without 21 individualized evidence. In re Nat'l Prescription Opiate Litig., 589 F. Supp. 3d at 808. The court squarely 22 8 Schall v. Suzuki Motor of Am., Inc., 450 F. Supp. 3d 771, 784 (W.D. Ky. 2020) ("Causation may be 23 proved by circumstantial evidence"); Mason v. Arizona Pub. Serv. Co., 622 P.2d 493, 500 (Ariz. Ct. App. 24 1980) ("Proximate cause may be determined from circumstantial evidence."); J.T. Baggerly v. CSX Transp., Inc., 635 S.E.2d 97, 101 (S.C. 2006) ("Normally, proximate cause is a question of fact for the 25 jury, and it may be proved by direct or circumstantial evidence."); Furlong v. Dyal, 539 S.E.2d 836, 838 (Ga. Ct. App. 2000) (plaintiff permitted to rely on "circumstantial and opinion evidence to create a jury 26 question" "on the issue of causation"); Rowhouses, Inc. v. Smith, 133 A.3d 1054, 1067 (Md. 2016) (circumstantial evidence and expert testimony can suffice to prove causation); Underwood v. Camden 27 Cnty. Off. of Sheriff, 2024 WL 1366672, at *3 (D.N.J. Mar. 28, 2024) ("circumstantial evidence can be 28 used to establish causation"). 189 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 202 of 235 1 rejected that argument, holding it "ignore[d] the aggregate nature of the evidence presented at trial and the 2 natural inferences allowed therefrom." Id. The aggregate evidence "supported a reasonable inference that 3 [each] Defendants' conduct was a substantial factor in creating the alleged nuisance." Id. The court 4 concluded, "aggregate proof of causation [is] sufficient to overcome summary judgment." In re Nat'l 5 Prescription Opiate Litig., 2019 WL 4178617, at *3. 6 Similarly, the Districts have offered aggregate evidence showing that each Defendant's conduct 7 was a substantial, concurring, or contributing factor in causing the Districts' harm. That is all the law 8 requires. See Bailey v. N. Am. Refractories Co., 95 S.W.3d 868, 871-73 (Ky. Ct. App. 2001) (applying 9 substantial factor test under Kentucky law); Stollenwerk v. Tri-W. Health Care All., 254 F. App'x 664 (9th 10 Cir. 2007) (applying substantial factor test under Arizona law); Gilbert v. Stewart, 255 A.3d 1101, 1114 11 (N.J. 2021) (applying substantial factor test under New Jersey law); Yonce v. SmithKline Beecham Clinical 12 Lab'ys, Inc., 680 A.2d 569, 576-77 (Md. Ct. Sp. App. 1996) (applying substantial factor test under 13 Maryland law); J.T. Baggerly, 635 S.E.2d at 101 (explaining that causation is satisfied where the evidence 14 establishes that the defendant's negligence is "a concurring or a contributing proximate cause.") (citation 15 omitted); Hayes v. Crawford, 730 S.E.2d 26, 29 (Ga. Ct. App. 2012) (similarly providing for concurrent 16 proximate causation). 17 2. Defendants' "entanglement" arguments on causation are premised on legal errors 18 a) Defendants misapprehend the features at issue by writing this Court's 19 failure to warn ruling out of existence. 20 To prevail on their failure to warn claims, the Districts are not required to prove that each 21 individual platform feature independently caused or contributed to their harm. This Court has repeatedly 22 held that Defendants may be held liable for failing to warn "of known risks of addiction attendant to any 23 platform features or as to platform construction in general." MTD Order, 754 F. Supp. 3d at 963–64 24 (emphasis added); accord In re Soc. Media Adolescent Addiction/Pers. Inj. Prods. Liab. Litig., 753 F. 25 Supp. 3d 849, 865 (N.D. Cal. 2024) ("AG MTD Order"). Defendants' suggestion that summary judgment 26 is warranted because the District have not linked each specific feature to the harms alleged ignores—and 27 would nullify—those rulings. In any event, the record contains ample evidence trying particular platform 28 features to the harms suffered. See supra at § III.A.2. 190 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 203 of 235 1 b) The Districts need not rule out every possible alternative causal factor to avoid Section 230. 2 Similarly, Section 230 also does not bar the Districts' claims simply because their harms are 3 factually related to "third-party wrongdoing or the publication of content." Mot. 16. Defendants' attempt 4 to graft a "but-for" causation test onto Section 230 has been squarely rejected by the Ninth Circuit. 5 Defendants misread Lemmon v. Snap, Inc., which held that "[t]he duty to design a reasonably safe 6 product is fully independent of Snap's role in monitoring or publishing third-party content." 995 F.3d 7 1085, 1093 (9th Cir. 2021). In holding Section 230 did not bar claims premised on Snap's product- 8 designer duties, the Ninth Circuit expressly acknowledged that "publishing content is a but-for cause of 9 just about everything Snap is involved in," yet clarified that this fact "does not mean that" holding Snap 10 liable for harm caused by its speed filter "seeks to hold Snap responsible in its capacity as a 'publisher or 11 speaker.'" Id. Even though content may be involved in Snap's operations—including, of course, the speed 12 filter, which was created to apply to users' person-to-person snaps—"that does not mean that" holding 13 Snap liable for harm caused by the speed filter would be holding it responsible as a publisher. Id. Section 14 230 protection only attaches where a plaintiff's claim would require Snap to "alter[] the content that 15 Snapchat's users generate" which is not being requested here. Id. at 1092; see id. ("Snap's alleged duty in 16 this case thus 'has nothing to do with' its editing, monitoring, or removing of the content that its users 17 generate through Snapchat."). 18 The rule from Lemmon is clear: under Section 230, a defendant may be held liable for its own 19 design decisions—even when those decisions affect how it distributes third-party content—but it cannot 20 be held liable for publishing or failing to remove that content. Yet in Defendants' telling, Lemmon imposes 21 an unprecedented limitation on tort liability, allowing claims to proceed only if "fully independent" of any 22 publishing activity. That is not what Lemmon said. The Ninth Circuit simply recognized that social media 23 companies have a duty to design safe platforms—separate from their editorial choices about what content 24 to publish. It did not hold that content must be "fully independent" of the causal chain. Indeed, if 25 Defendants' selective reading were correct, Lemmon itself would have been dismissed—since the alleged 26 harm there arose from Snap's Speed Filter, a feature inherently intertwined with user content. See 995 27 28 191 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 204 of 235 1 F.3d at 1092 (recognizing that "Snap acted as the 'publisher or speaker' of user content by transmitting 2 Landen's snap"). That is plainly not what Lemmon held, no matter how Defendants try to spin it. 3 At any rate, Defendants' attempt to reintroduce a "but for" test into Lemmon has been rejected by 4 subsequent Ninth Circuit precedent. The court has clarified that Section 230 applies only when a plaintiff 5 seeks to hold a defendant liable for third-party content, not when a claim challenges the company's own 6 design choices, even if those choices relate to such content. In Bride v. Yolo Technologies, Inc., the court 7 explained that the claim in Lemmon "did not depend on third-party content"—and Section 230 therefore 8 did not apply—because it "did 'not depend on what messages, if any, a Snapchat user employing the 9 Speed Filter actually sends.'" 112 F.4th 1168, 1180 (9th Cir. 2024) (quoting Lemmon, 995 F.3d at 1094). 10 In Doe 1 v. Twitter, 148 F.4th 635 (9th Cir. 2025), the court reaffirmed that Lemmon involved treating 11 Snap as a "'product designer' rather than a 'publisher or speaker,'" because the claim "turned on Snap's 12 design architecture rather than the publication of any content." Id. at 645. The Doe 1 plaintiffs alleged 13 defects in Twitter's mechanism for reporting CSAM—defects inseparable from the existence of third- 14 party content. Yet Section 230 still did not apply, because the claim targeted Twitter's defective "reporting 15 mechanism," which could be fixed "without monitoring, removing, or in any way engaging with third- 16 party content." Id. 17 Calise v. Meta Platforms, Inc. underscores the same principle. In that case, the Ninth Circuit 18 explained, "it is not enough that a claim, including its underlying facts, stems from third-party content 19 for § 230 immunity to apply." 103 F.4th 732, 742 (9th Cir. 2024) (emphasis added). Indeed, the court 20 recognized that nearly any claim alleging a failure to warn about an online platform could, in some sense, 21 "concern content." Id. at 742. But that alone is not enough to trigger Section 230. As Calise observed, "it 22 is not true that 'providing a warning' would not require 'considering the content posted.' How could 23 Internet Brands warn about certain harmful content without considering what it was? Such a rudimentary 24 fact-bound inquiry quickly falls apart and runs up against our precedent." Id. Section 230 was never 25 intended to bar claims simply because content exists in the background. Sometimes content is involved, 26 and that's fine. 27 In the bellwether cases, third-party content is related to various defective design features of 28 Defendants' platforms. But, importantly, the Districts do not allege harm from the content itself—and, as 192 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 205 of 235 1 school districts rather than individual users, they do not allege having viewed content at all. Under these 2 circumstances, Section 230 is plainly inapplicable. 3 The additional cases Defendants cite do not help them. In M.P. ex rel. Pinckney v. Meta Platforms 4 Inc., 127 F.4th 516 (4th Cir. 2025), the Fourth Circuit applied Section 230 because the plaintiff sought to 5 "hold Facebook liable for disseminating 'improper content' on its website." Id. at 525. Likewise, in Doe 6 (K.B.) v. Backpage.com, LLC, 724 F. Supp. 3d 882 (N.D. Cal. 2024), Section 230 applied because the 7 plaintiff's "entire claim [was] based on the content of the subsequent interaction" with a third party." Id. 8 at 885. Nuveen Mun. High Income Opportunity Fund v. City of Alameda, 730 F.3d 1111, 1121 (9th Cir. 9 2013) and In re Williams Sec. Litig.-WCG Subclass, 558 F.3d 1130, 1138 (10th Cir. 2009) addressed loss 10 causation in securities fraud, not factual or proximate causation in tort. Here, by contrast, the record 11 contains ample evidence creating triable issues of causation. 12 In reply, Defendants may also point to Doe v. Grindr Inc., 128 F.4th 1148, 1153 (9th Cir. 2025), 13 but that case does not support their expansive view of Section 230. There, the plaintiff challenged Grindr's 14 decision to allow him "to communicate with abusive adults." Id. In Grindr, the communication itself was 15 the harm—the wrongful act and injury were one and the same. The claim thus turned on third-party content 16 and publishing activity; the alleged harm and content were inseparable, not "independent" of each other. 17 See id. By contrast, the Districts' claims do not arise from any third-party communication or content, but 18 from Defendants' own defective platform design and failure to warn of known risks. See id. 19 c) Juries are routinely called upon to evaluate liability in the context of concurrent causation 20 21 Defendants are also wrong as a matter of basic tort law: They are not entitled to summary judgment 22 simply because the Districts cannot perfectly isolate the harm caused by each platform. Under settled 23 common law, "a tortfeasor is liable for any injury of which his negligence is a proximate cause." Am. 24 Motorcycle Ass'n v. Super. Ct., 578 P.2d 899, 904 (Cal. 1978). Even if "it is simply impossible to 25 determine whether or not a particular concurrent tortfeasor's negligence, acting alone, would have caused 26 the same injury," each may still be held liable for the harm it proximately caused. Id. at 589. 27 Left unsaid in Defendants' motions is that every bellwether jurisdiction applies the same rule: 28 proximate cause exists where a defendant's "conduct was a substantial factor in bringing about the injuries, 193 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 206 of 235 1 even where there are other intervening causes which were foreseeable or were normal incidents of the risk 2 created." Gilbert v. Stewart, 255 A.3d 1101, 1114 (N.J. 2021). This standard "does not require an 3 unsevered connecting link between the negligent conduct and the ultimate harm." Id. (emphasis omitted). 4 Rather, a plaintiff need only show it is "reasonably probable" that each defendant's conduct contributed 5 to its injuries. E.g., Walton v. Premier Soccer Club, Inc., 334 A.3d 784, 796–97 (Md. 2025).9 The Districts 6 survive summary judgment on this ground because they have established it is "more likely than not" each 7 Defendant's platform contributed to their alleged harm. Walton, 334 A.3d at 793. It would defy logic and 8 settled tort law to hold that each Defendant may be a "substantial" factor in causing harm, yet escape 9 liability simply because their misconduct is intertwined with that of other tortfeasors. 10 These basic legal principles also resolve Defendants' argument that the Districts must completely 11 isolate the harm caused by their defective features from any potential negative effects from third-party 12 misconduct, or even other platform features Defendants claim are not at issue (despite this Court's failure- 13 to-warn ruling). The existence of "preexisting social issues affecting the community" does not undermine 14 the Districts' claims, and the Districts are not required "to precisely apportion [their] injuries" to the 15 alleged harm rather than to unrelated societal factors. Ramah Navajo Sch. Bd., Inc. v. Sebelius, 2013 WL 16 12303945, at *19 (D.N.M. May 9, 2013). Instead, "it is the duty of the trier of fact to make the best 17 estimate in its power to base its damages calculation on the actionable injury itself." Id.; see Restatement 18 (Second) of Torts § 433A(1) cmt. b (where there are multiple concurrent causes, damages should be 19 awarded if "it is possible to make a rough estimate which will fairly apportion such subsidiary elements 20 of damages."); see also, e.g., CRS Sirrine, Inc. v. Dravo Corp., 464 S.E.2d 897, 900 (Ga. Ct. App. 1995) 21 9 22 The other bellwether states' laws are in accord. See, e.g., Mayer v. Hous. Auth. of Jersey City, 202 A.2d 439, 447 (N.J. Super. Ct. App. Div. 1964), aff'd 210 A.2d 617 (N.J. 1965) ("If the evidence, viewed in the 23 light most favorable to plaintiff, would cause fair-minded men to differ as to whether there was a reasonably probable relation of cause and effect between the alleged negligence and the injuries, the issue 24 must be submitted to the jury."); Robertson v. Sixpence Inns of Am., Inc., 789 P.2d 1040, 1046–47 (Ariz. 1990) ("Plaintiff need only present probable facts from which the causal relationship reasonably may be 25 inferred."); Barrett Props., LLC v. Roberts Capitol, Inc., 729 S.E.2d 621, 624 (Ga. Ct. App. 2012) ("To establish proximate cause, a plaintiff….must introduce evidence which affords a reasonable basis for the 26 conclusion that it is more likely than not that the conduct of the defendant was a cause in fact of the result."); accord Thacker v. Ethicon, Inc., 47 F.4th 451, 460 (6th Cir. 2022); Allen v. Greenville Hotel 27 Partners, Inc., 2006 WL 1817804, at *6 (D.S.C. June 30, 2006) (quoting Ballou v. Sigma Nu Gen. Fraternity, 352 S.E.2d 488, 493 (S.C. Ct. App. 1986)) (In South Carolina, causation can only be resolved 28 on a motion for summary judgment in "rare or exceptional cases."). 194 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 207 of 235 1 ("if a plaintiff can show with reasonable certainty the total amount of damages and the degree to which 2 those damages are attributable to defendant, that is sufficient to support an award."). Aside from failure- 3 to-warn, evidence concerning features shielded by Section 230 remains relevant for context.10 4 C. Genuine material disputes of fact concerning the Districts' damages preclude summary judgment. 5 1. Each Plaintiff has presented evidence of compensable "hard costs." 6 The Court ruled that school districts may seek damages for various costs incurred as a result of 7 Defendants' negligence and nuisance, including but not limited to: "[1] diverting and increasing financial 8 resources to address the disruptive forces of defendants' social media products in school; [2] hiring mental 9 health personnel and developing mental health resources; [3] implementing new information technology 10 and physical resources to limit access to and mitigate risks caused by defendants' platforms; and [4] 11 repairing property damage." MTD Order, 754 F. Supp. 3d at 967. Consistent with the Court's ruling, the 12 Districts have supplied invoices detailing their costs, which were described in the expert reports of Jeffrey 13 Meyers. See Exs. 574, 575, 576, 577, 578, 579. District employees have likewise attested that these costs 14 directly resulted from Defendants' misconduct. See infra § III.C.1.a. 15 Defendants do not claim that the Districts' hard costs are not recoverable as a matter of law and, 16 with limited exceptions, do not challenge whether the Districts have evidence of those costs. Instead, they 17 claim that the schools have no "competent" evidence of past damages. Defs.' Tucson Mot. at 25; Defs.' 18 Dekalb Mot. at 22; Defs.' Breathitt Mot. at 20; Defs.' Irvington Mot. at 21; Defs.'Charleston Mot. at 23. 19 This is shorthand for asking the Court to cut out the jury and weigh the evidence, which is impermissible 20 at summary judgment. See Berman v. Freedom Fin. Network, LLC, 400 F. Supp. 3d 964, 969 (N.D. Cal. 21 2019) (Gonzalez-Rogers, J.) ("the weighing of the evidence, and the drawing of legitimate inferences from 22 facts are jury functions, not those of a judge."). 23 Because the schools have presented evidence that their costs were incurred due to Defendants' 24 misconduct, the question is for the jury to weigh and make an ultimate finding about causation. In every 25 bellwether state, questions of general and proximate cause are for the jury unless there is no possibility of 26 27 10 28 The Districts more fully respond to this misguided argument in the 230 Daubert Opp. 195 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 208 of 235 1 an inference supporting the non-moving party. Torres v. Jai Dining Servs. (Phoenix) Inc., 497 P.3d 481, 2 486 (Ariz. 2021); Johnson v. Wood, 2025 WL 2715888, at *4 (Ga. Ct. App. Sept. 24, 2025) ("Questions 3 of negligence, diligence, contributory negligence, and proximate cause are peculiarly matters for the jury, 4 and a court should not take the place of the jury in solving them, except in plain and indisputable cases." 5 (citation omitted)); Pathways, Inc. v. Hammons, 113 S.W.3d 85, 89 (Ky. 2003); Pittway Corp. v. Collins, 6 973 A.2d 771, 253 (Md. Ct. App. 2009); Gilbert v. Stewart, 255 A.3d 1101, 1113–14 (N.J. 2021); 7 Wickersham v. Ford Motor Co., 853 S.E.2d 329, 390–91 (S.C. 2020). This is not one of those rare cases. 8 The record contains ample evidence of causation, which remains disputed and must therefore be resolved 9 by the jury. Defendants can argue to the jury that these costs were attributable to some other cause, and 10 the jury will evaluate and weigh the evidence presented. 11 a) The Districts presented evidence of incurred costs relating to mental health resources, and technology materials, systems, and 12 programming. 13 The Districts have adduced evidence of costs incurred to address Defendants' misconduct, such as 14 costs of adding mental health resources and implementing new technology, including physical materials, 15 systems, and programming. The Court has already confirmed that such categories of damages are 16 recoverable. MTD Order, 754 F. Supp. 3d at 963. 17 Technology Lockers. Several Districts seek compensation for purchasing physical technology 18 lockers (i.e. Yondr pouches11, Amazon Capital lockers, or other similar devices) in response to 19 Defendants' conduct.12 While Defendants object to these costs by arguing the lockers are intended to stop 20 students "from accessing their phones generally, not just Defendants' platforms," the Districts have 21 22 11 23 Yondr pouches are locking magnetic pouches purchased by schools to secure students' phones during the school day, limit cell phone and social media use, reduce classroom distractions, and promote a more 24 focused learning environment. 12 25 Technology Lockers. Tucson: Yondr pouches ($16,850) and Amazon Capital lockers ($16,101). See Ex. 579 at App'x C Ex. 1 at 1-2. DeKalb: Yondr pouches ($391,076) and McGee phone lockers 26 ($21,280). See Ex. 576 at App'x C Ex. 1 at 1. Breathitt: Quill cellphone caddies ($209) and Amazon Capital lockers ($1,758). See Ex. 574 at App'x C Ex. 1 at 1. Charleston: Yondr pouches ($273,953). 27 See Ex. 575 at App'x C Ex. 1 at 1. Harford and Irvington do not seek compensation for technology 28 lockers. 196 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 209 of 235 1 presented evidence that the lockers would not have been purchased but for social media use in schools.13 2 Defendants do not engage with these proffered facts, which create a clear question for the jury to determine 3 whether social media use from Defendants' platforms was a substantial factor in requiring the purchase 4 of technology lockers. That there may be other benefits of locking phones away during school is not 5 dispositive, and Defendants present no evidence that other cellphone use is the sole or superseding cause 6 necessitating locker purchases. 7 Filtering and Monitoring Software. Certain Districts seek partial compensation for purchases of 8 internet filter systems and school software that block social media websites in schools or assist school 9 staff in monitoring students' use: 10 • Breathitt: 75% compensation for implementing monitoring software Hapara ($47,468); 50% compensation for implementing E-Hallpass software Eduspire ($725); and 25% compensation 11 for implementing E-Hallpass software Securly ($2,222). See Ex. 574 at App'x C Ex. 1 at 1-3. • Tucson: 40% compensation for implementing online reporting system Awareity ($33,582). 12 See Ex. 579 at App'x C Ex. 1 at 1. 13 • Irvington: 35% compensation for implementing filtering systems GoGuardian ($73,185) and Palo Alto ($47,565). See Ex. 578 at App'x C Ex. 1 at 5-6. 14 • DeKalb: 30% compensation for implementing filtering system Lightspeed ($1,173,042). See Ex. 576 at App'x C Ex. 1 at 1. 15 • Harford: 20% compensation for implementing filtering systems Skyline ($68,447) and Cisco 16 Umbrella ($64,346). See Ex. 577 at App'x C Ex. 1 at 1. • Charleston does not seek compensation for internet filtering systems or monitoring software. 17 18 Defendants argue that these systems are not used solely to block or respond to social media, (e.g. Defs.' 19 Tuscon Mot. at 34; Defs.' Breathitt Mot. at 25), or that schools may have been required to implement 20 filtering software irrespective of Defendants' platforms (Defs.' Dekalb Mot. at 28; Defs.' Harford Mot. at 21 31-32; Defs.' Irvington Mot. at 27). However, the Districts have presented evidence that these costs are 22 23 24 25 13 Tucson: Ex. 1125 at 41:19-44:3; Ex. 1126 ¶ 12 (Sabino installed cellphone lockers to physically remove 26 phones and access to social media from students because alerts prompt checking social media instead of focusing on class); DeKalb: Ex. 1127 at 82:10-83:6 (implementation of pouches and lockers was a direct 27 response to distractions from social media platforms); Breathitt: Ex. 1128 at 16:11-15 (district would not 28 have purchased cell phone caddies "if it was not for social media"); Charleston: Ex. 1129 at 58:1-22. 197 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 210 of 235 1 in part attributable to Defendants' platforms, so these facts are in dispute.14 2 Mental Health Resources. Certain Districts seek compensation for new and increased mental 3 health resources, as set out below. The District adduce evidence that Defendants' platforms were a 4 substantial factor in necessitating expansion of these resources.15 5 • Tucson: 40% compensation for access to online therapy platform Talkspace ($231,021). See Ex. 579 at App'x C Ex. 1 at 2. 6 • Irvington: 20% compensation for mental health resources through New Jersey Coalition for Inclusive Education ($325,147), Care Plus NJ ($624,917), Brett Dinovi & Associates 7 ($226,467), Live Breathe Calm ($19,770), Generations Family Guidance ($58,160), Center for Partnership Services ($146,136), and Momentum Therapy Services ($66,086). See Ex. 578 at 8 App'x C Ex. 1 at 1-7. 9 • Breathitt, Charleston, DeKalb, and Harford and do not seek compensation for expanded mental health resources beyond the lost time for staff providing those resources, as discussed 10 in the lost time section. 11 Defendants claim that the estimated percentage compensation for Tucson's resources lacks sufficient 12 basis, see Defs.' Tucson Mot. at 34-35, and that one of Irvington's vendors charges a flat rate that it would 13 have incurred absent Defendants' misconduct. See Defs.' Irvington Mot. at 27. But see Ex.579; Ex. 582; 14 Ex. 668. These arguments go to the weight of the evidence and are for the jury to consider. Defendants 15 admit that the District experts relied on sworn interrogatory responses that confirm these resources were 16 required in part due to social media use in schools. The fact that these programs may provide benefit to 17 some students beyond the context of social media use does not erase the fact that they were used in 18 19 14 Breathitt: Ex. 661 at 22:11-23:9 (Hapara's focus is "keeping [students] off their social media and then 20 also keeping them off videos, YouTube videos"), 23:16-24 (Hapara "was purchased above and beyond just the normal blocking and filters."); Ex. 1133 at 128:11–22 (Breathitt uses Hapara to disconnect 21 students from social media); DeKalb: Ex. 1130 at 143:2–24 (Lightspeed is designed to block access to 22 social media websites); Ex. 520 ¶¶ 8-15 ("30% of Lightspeed costs are related to the District's attempts to block and/or limit social media."); Harford: Ex. 598 at 28:6-25 (explaining 20% attribution of costs to 23 social media events), 21:5-24:25 (describing filtering specific to social media, not just what law requires); Ex. 571 at 13:11 (describing that Cisco and Skyline are used to block social media). 24 15 Tucson: Ex 543 at 72:12-17 (district has seen an "increase of anxiety and depression and overall mental 25 health supports needed on [its] campuses," with conditions "heavily influenced by social media"), 73:4- 16 (increased "anxiety, depression, and body dysmorphia" attributable to "social media platforms" drove 26 need for increased mental health resources); Irvington: Ex. 524 at 99:22-100:8 (describing social media as a "main" reason why increased mental health resources were required); Ex. 600 at 146:1-147:16 27 (attributing social media as a substantial factor in mental health harms for IPS students, including by 28 exacerbating preexisting problems). 198 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 211 of 235 1 response to Defendants' misconduct. Defendants offer no evidence that an alternate cause was the sole or 2 superseding cause of these costs, and even if they did, that would be in dispute. 3 Programming. The same is true of evidence of new and increased social emotional learning (SEL) 4 programming costs and other responsive costs, which the Districts have incurred as follows: 5 • Breathitt: 50% compensation for programming from Kids First ($498); 25% compensation for programming from Remix Education ($2,733); 50% compensation for programming from 6 Millstone Labs ($6,719). See Ex. 574 at App'x C Ex. 1 at 1-2. • Tucson: 40% compensation for programming with Character Strong ($242,856). See Ex. 579 7 at App'x C Ex. 1 at 1-2. 40% compensation for creating Comprehensive School Threat Assessment Guidelines ($2,912). See Ex. 579 at App'x C Ex. 1 at 2. 8 • Charleston: 7% compensation for programming from Flippen Group ($95,476); 20% 9 compensation for implementing a social emotional learning platform from Panorama Education ($192,105); 40% compensation for restorative practices training and coaching from 10 International Institute for Restorative Practices ($163,278); 35% compensation for social emotional learning curricula from Committee for Children for the Second Step ($226,278); 11 40% compensation for programming from Restorative Resolutions ($33,110); 40% compensation for educational training services provided by Restorative Coaching ($9,600); 12 80% compensation for educational services from Social Emotional Learning Alliance for South Carolina ($9,600). See Ex. 575 at App'x C Ex. 1 at 1-9. 13 • DeKalb, Harford, and Irvington are not seeking compensation for hard costs of expanded programming. 14 15 The Districts have presented evidence that they implemented various programming in part as a response 16 to rising and disruptive social media use in schools.16 It is for the jury to determine whether the evidence 17 establishes that Defendants were a substantial factor in causing the Districts to incur those costs. 18 b) Defendants' cross-cutting entanglement arguments do not support summary judgment on causation 19 20 16 21 Tucson: Ex. 543B at 354:10-15; see also id. at 357:21-358:6 (connecting the need for social emotional curriculum in Tucson to the negative impact of social media on students' social emotional well-being), 22 362:10-19 (noting that Character Strong curriculum is focused on social emotional skills and includes specific lessons about social media); Ex. 543A at 71:6-72:17 (increased anxiety, depression, and overall 23 mental health supports needed by students heavily influenced by Defendants' platforms); Ex. 547 at 24 188:22-190:5 (identifying substantial work with SEL curriculum to help address students' needs resulting from student use of social media, specifically TikTok, Instagram, YouTube, and Snapchat); Breathitt: 25 Ex. 1128 at 46:23–47:14 (Cyber Safe Week to discuss dangers of social media apps); Charleston: Ex. 548 ¶ 31 (Screenagers directly attributable to Ds' conduct); id. ¶ 26 (increased use of Panorama attributed 26 to Ds' conduct); id. ¶ 24; id. ¶ 25 (connecting Flippen Group's Capturing Kids' Hearts to social media); Ex. 566B at 200:15-202:16 (apportioned for costs working with adults rather than students); id. at 64:15- 27 95:22 (connecting Second Step to alleged harms); Ex. 1186 at 88:21-90:10 (connecting restorative 28 practices curriculum to social media). 199 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 212 of 235 1 Defendants argue that the Districts failed to attribute their costs solely to Defendants' conduct. But 2 this disregards a "long-recognized tenet of tort law" that there can be multiple "'but-for' causes" that 3 support liability, and as a result a plaintiff is not required to show actionable conduct "was the 'sole' cause 4 at summary judgment." See Moore v. City of Atlanta, Ga., 2022 WL 19517299, at *25 (N.D. Ga. Dec. 19, 5 2022); supra § III.B.2. 6 Along the same lines, Defendants argue that the Districts' costs are too difficult to calculate or 7 apportion given multiple potential causes. That is no basis to grant summary judgment on causation. Under 8 the law of all bellwether states, apportionment of damages among defendants is a question for the jury 9 that flows from its determination of fault for any (or all) of the Defendants. Piner v. Sup. Ct. in and for 10 Cnty. of Maricopa, 962 P.2d 909, 915 (Ariz. 1998) (defendant's liability is not "limited by apportioning 11 damages, but only by apportioning fault"); Quynn v. Hulsey, 850 S.E.2d 725, 729 (Ga. 2020) (jury 12 apportions damages between multiple tortfeasors "according to the percentage of their fault"); Fed. 13 Deposit Ins. Corp. v. Loudermilk, 826 S.E.2d 116, 126 (Ga. 2019) (same); Owens Corning Fiberglas 14 Corp. v. Parrish, 58 S.W.3d 467, 479 (Ky. 2001) ("If . . . the evidence does not permit apportionment of 15 the damage between separate causes, then comparative fault principles apply, and the trial court should 16 instruct the jury to apportion damages according to the proportionate fault of the parties."); Carter v. 17 Wallace & Gale Asbestos Settlement Tr., 96 A.3d 147, 160 (Md. Ct. App. 2014) (noting that where injury 18 is divisible and apportionment is a fact question, burden rests with defendant to show how damages should 19 be apportioned); Campione v. Soden, 695 A.2d 1364, 1370-71 (N.J. 1997) (discussing apportionment and 20 reasoning that the question of apportionment is for the jury); Pratt v. Amisub of SC, Inc., 912 S.E.2d 268, 21 282-83 (S.C. Ct. App. 2025) (explaining South Carolina's joint tortfeasor statutory scheme, including role 22 of jury in apportioning fault and damages); see also Restatement (Second) of Torts § 433A (explaining 23 difference between apportionment for divisible injuries with an evidentiary basis for determining 24 contribution, versus indivisible harms where evidence that guides apportionment of damages is lacking). 25 Defendants separately criticize the Districts for relying on interrogatory responses. E.g., Defs.' 26 Tucson Mot. at 33, 35. These criticisms are unavailing. Experts are permitted to rely upon sworn 27 interrogatory responses, and Rule 56 expressly contemplates that interrogatories may be used to 28 demonstrate a genuine dispute of material fact that will defeat a motion for summary judgment. See 200 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 213 of 235 1 Correct Transmission, LLC v. Nokia of Am. Corp., 2024 WL 1289821, at *5 (E.D. Tex. Mar. 26, 2024); 2 Fed. R. Civ. P. 56(c)(1)(A). If Defendants wish to challenge the sufficiency of Plaintiffs' interrogatory 3 responses, which have been verified under penalty of perjury, they can present that argument to the jury. 4 2. Defendants can be required to compensate Districts for the diversion of 5 school resources caused by their misconduct. 6 Defendants' challenge to the Districts' diversion-of-resources damages fails for multiple, 7 independently sufficient reasons. Across their six motions for summary judgment, Defendants 8 fundamentally misstate the law in asserting that the Districts cannot recover the costs of employee time 9 diverted to address the disruptions to instructional time and school operations caused by Defendants' 10 social media platforms. Defendants' position rests on inapposite authority and a mischaracterization of 11 compensable damages, improperly suggesting that the Districts may recover only new or additional 12 expenditures incurred as a direct result of Defendants' misconduct. The law is clear that diversion of 13 resources itself constitutes a recognized and compensable injury, and courts have long permitted recovery 14 of such damages. See Convoy Co. v. Sperry Rand Corp., 672 F.2d 781, 785 (9th Cir. 1982). 15 As an initial matter, the Ninth Circuit has squarely rejected Defendants' improper contention that 16 the Districts cannot recover for employee time lost to addressing harms caused by Defendants, simply 17 because the Districts would have paid those employees' salaries and benefits regardless. In Convoy Co. v. 18 Sperry Rand Corp., the plaintiff sought damages for "the hours its salaried personnel spent supervising [a 19 malfunctioning] computer system." Id. at 783. On appeal, the defendant argued that such damages were 20 impermissible because "salaried supervisory staff costs . . . [are] not allowable as a matter of law" and the 21 plaintiff "would have paid the staff's salary in any event." Id. at 785. The Ninth Circuit expressly rejected 22 that argument, holding that "the issue is not whether [the plaintiff] would have paid the supervisors' 23 salaries if the defendant had not breached the contract, but whether the breach deprived Convoy of the 24 services it paid for." Id. 25 Indeed, "[m]ost courts that have considered the issue have concluded that a plaintiff can recover 26 the value of employees' lost services as damages in a contract or tort action, even when it had not shown 27 that it incurred additional expenses or lost profits." State v. Rouse, 647 N.W.2d 286, 289 (Wis. Ct. App. 28 2002) (emphasis added) (collecting cases); see also Zayo Grp., LLC v. Sols. Fiber Optic, Inc., 2025 WL 201 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 214 of 235 1 2157902, at *4 (E.D. Va. July 11, 2025), report and recommendation adopted, 2025 WL 2155792 (E.D. 2 Va. July 29, 2025) (ordering damages including "internal labor costs"); Mobile Conversions, Inc. v. 3 Allegheny Ford Truck Sales, 2014 WL 7369898, at *6 (W.D. Pa. Dec. 29, 2014); Dunn Appraisal Co. v. 4 Honeywell Info. Sys. Inc., 687 F.2d 877, 883 (6th Cir. 1982) (upholding award of 1/5 of a CEO's salary 5 based on time spent addressing fallout from defendants' conduct). 6 Specifically, Kentucky courts recognize that lost time and opportunity costs incurred while 7 mitigating the consequences of another's wrongful conduct constitute a cognizable form of damages in 8 negligence actions. See Lurry v. PharMerica Corp., 2024 WL 2965642, at 3 (W.D. Ky. June 12, 2024) 9 ("lost time and opportunity costs associated with attempting to mitigate the actual consequences of the 10 defendant's conduct" are recoverable damages); see also Shulz v. Chadwell, 558 S.W.2d 183, 187 (Ky. 11 Ct. App. 1977) (holding that where one partner became unavailable for work, the amount paid to the other 12 partner to perform the same work was "sufficient evidence of probative value on the issue of lost time"). 13 Defendants' cited authorities do not support a contrary conclusion. See Schwartz v. Hasty, 175 S.W.3d 14 621, 625 (Ky. Ct. App. 2005) (not addressing lost employee time); Gassaway Constr. Co. v. Gentry, 264 15 S.W.2d 658, 659 (Ky. 1954) (finding insufficient evidence to establish loss-of-time damages). 16 Similarly, Georgia courts permit plaintiffs to seek recovery of portions of employee salaries, such 17 as those sought here, and reject arguments that permitting such damages would allow recovery of costs 18 the plaintiff would have otherwise incurred. See Aaron's, Inc. v. MDC Grp., Inc., 2010 WL 11505919, at 19 *7 (N.D. Ga. Dec. 15, 2010). Similarly, the Third Circuit, applying New Jersey law, upheld an award for 20 compensatory damages including "the value of the time [Plaintiff's] employees expended in attempting 21 to reduce or avert the damage flowing from the" defendant's conduct. Comdyne I, Inc. v. Corbin, 908 F.2d 22 1142, 1146, 1150 (3d Cir. 1990). Courts in Arizona have also upheld damages seeking compensation for 23 lost employee time. See PivotHealth Holdings LLC v. Horton, 2025 WL 1865788, at *2 (D. Ariz. July 7, 24 2025) (finding standing based on damages including "employee time" that was "incurred in responding 25 to" the defendant's conduct); IceMOS Tech. Corp. v. Omron Corp., 2020 WL 1083817, at *4 (D. Ariz. 26 Mar. 6, 2020) (permitting plaintiff to present evidence at trial "seeking costs of the salaries of its 27 employees for time spent working" to address defendant's conduct). 28 Defendants' own cited authority confirms that lost employee time is a compensable form of 202 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 215 of 235 1 damages under South Carolina law. See Charleston Lumber Co. v. Miller Hous. Corp., 458 S.E.2d 431, 2 437 (S.C. Ct. App. 1995) ("The Millers alleged damages of lost employee time based on their personnel 3 having to spend numerous hours each month checking and correcting the bids versus the actual charges. 4 We find that employee time is a compensable damage."). While Defendants suggest -- without citing any 5 authority—that Charleston Lumber's holding is limited to fraud claims, courts routinely recognize lost- 6 time damages in negligence actions. See In re Blackbaud, Inc., Customer Data Breach Litig., 567 F. Supp. 7 3d 667, 687 (D.S.C. 2021) (finding plaintiffs alleged cognizable injury under South Carolina law for time 8 and money spent mitigating injuries arising from defendant's negligence). 9 Finally, courts applying Maryland law have expressly rejected the same argument Defendants 10 advance here—that no loss is incurred because an employer would have paid the employees anyway. See 11 Under Armour, Inc. v. Ziger/Snead, LLP, 158 A.3d 1134, 1138 (Md. Ct. Spec. App. 2017) (holding that 12 "diverted employee time [is] a compensable loss"). As Maryland courts have explained, because "diverted 13 employee time" prevents employees from rendering "other service to the [plaintiff]," such costs are 14 recoverable. MacDonald v. Patriot, LLC, 2017 WL 1788115, at *12 (Md. Ct. Spec. App. May 5, 2017); 15 see also KeraLink Int'l, Inc. v. Stradis Healthcare, LLC, 2021 WL 5832242, at *1 (D. Md. Nov. 2, 2021), 16 aff'd sub nom. KeraLink Int'l, Inc. v. Geri-Care Pharms. Corp., 60 F.4th 175 (4th Cir. 2023) (entering 17 judgment for damages including "costs of employee time"). 18 Defendants rely on authority disallowing recovery for lost personal time. That authority is 19 inapposite: Those cases involved plaintiffs seeking compensation for their own inconvenience, not for the 20 loss of employee services suffered by an organizational plaintiff. See Quinalty v. FocusIT LLC, 2024 WL 21 342454, at *4 (D. Ariz. Jan. 30, 2024) (plaintiff alleged lost personal time responding to data breach); 22 Griffey v. Magellan Health Inc., 562 F. Supp. 3d 34, 45 (D. Ariz. 2021) (same); In re Gen. Motors LLC 23 Ignition Switch Litig., 339 F. Supp. 3d 262 (S.D.N.Y. 2018) (plaintiffs alleged lost personal time seeking 24 repair of their vehicle). The Districts' claim is different. The Districts do not seek recovery for personal 25 inconvenience, but for the value of employee time and services diverted from their intended educational 26 and operational functions to address the disruptions caused by Defendants' platforms. As the Sixth Circuit 27 recognized, "[employer] was paying [employee], and the time he spent was [employer]'s time, not his 28 own." Dunn Appraisal Co. v. Honeywell Info. Sys. Inc., 687 F.2d 877, 884 (6th Cir. 1982). 203 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 216 of 235 1 As Dr. Ward explained, his calculation of damages based on diverted employee time "is based on 2 the fundamental economic concept of opportunity cost," and the harm to the Districts is "the value of the 3 foregone alternative." E.g., Ex. 1083 ¶ 8. To quantify the cost of these foregone activities, "[e]conomists 4 often estimate the opportunity cost of time using the individual's wage." Id. ¶ 9. This is a widely-accepted 5 method of calculating the opportunity costs in an educational setting. As Dr. Ward explained, where an 6 intervention "requires teachers to reallocate their time from instruction, that time should be counted as an 7 opportunity cost to the school represented by the teachers' wages and benefits for that time." Ex. 1084 ¶ 8 17 (quoting Levin et al., Economic Evaluation in Education: Cost-Effectiveness and Benefit-Cost 9 Analysis, 4th ed. 2017, at 69). 10 In short, the diversion of teacher and staff time from their educational mission, to address the 11 disruptions caused by Defendants' social media platforms, constitutes a compensable form of harm, 12 recognized both by established precedent and economic principles. While Defendants fault the Districts 13 for responding to this novel and externally imposed harm with the resources available to them, state and 14 federal law uniformly recognize that when a defendant's conduct deprives a plaintiff of the services of its 15 personnel, that deprivation is a cognizable form of damages. See Convoy Co., 672 F.2d at 785. Defendants' 16 motions should be denied to the extent they seek to preclude the Districts' recovery for the substantial loss 17 of employee time and resources diverted to address the harms caused by Defendants' platforms. 18 3. Mr. Klein's survey raises disputed issues of fact regarding the Districts' lost 19 time damages. 20 Defendants incorrectly contend that the Districts lack evidence of their lost time damages by 21 challenging the teacher survey conducted by Mr. Klein. As explained in Plaintiffs' Omnibus Daubert 22 Opposition, Mr. Klein—an MIT-educated expert in market research with over fifty years of experience 23 and more than 1,000 surveys administered—designed and executed the survey in accordance with well- 24 accepted principles of market research. Its findings provide reliable, admissible evidence supporting the 25 Districts' damages and raise disputed issues of fact regarding lost instructional time. See SD Daubert Opp. 26 § III.B; Ex. 1085 ¶ 41 (Breathitt: scheduled classroom time diverted 1.4-5.2% for middle school, 5.2-3.9% 27 for high school); Ex. 1086 ¶ 41 (Charleston: scheduled classroom time diverted 5.5-7.1% for middle 28 school, 4.1-4.3% for high school); Ex. 1087 ¶ 41 (DeKalb: scheduled classroom time diverted 1.6-5.2% 204 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 217 of 235 1 for middle school, 4.6-11.7% for high school); Ex. 1088 ¶ 41 (Harford: scheduled classroom time diverted 2 1.9-5.7% for middle school, 3.1-10.6% for high school); Ex. 1089 ¶ 41 (Irvington: scheduled classroom 3 time diverted 4.7-2.2% for middle school, 3.7-9.9% for high school); Ex. 1090 ¶ 41 (Tucson: scheduled 4 classroom time diverted 2.6-5.2% for middle school, 3.7-14.8% for high school). 5 Mr. Klein's survey quantifies the instructional time lost because of Defendants' conduct based on 6 the accounts of those with first-hand knowledge—teachers and administrators directly impacted by 7 Defendants' platforms. While Defendants seek to dismiss these educator accounts, Mr. Klein's analysis 8 offers a scientifically valid and methodologically sound measure of the Districts' diverted resources and 9 lost time. See SD Daubert Opp. § III.B. Courts routinely recognize that survey evidence based on 10 employee self-reporting can provide a reasonable approximation of damages where detailed 11 contemporaneous records are unavailable. See, e.g., Alcantar v. Hobart Serv., 2013 WL 156530, at 4 (C.D. 12 Cal. Jan. 15, 2013) ("It would be unreasonable to expect employees to maintain records of every time they 13 were provided a meal break," and "survey evidence … is relevant to give an approximation of damages."). 14 Ultimately, Defendants' criticisms of the Klein survey go to weight, not admissibility. The survey 15 easily satisfies Rule 702 and creates genuine issues of material fact sufficient to preclude summary 16 judgment on the issue of lost time. See id.; Fortune Dynamic, Inc. v. Victoria's Secret Stores Brand Mgmt., 17 Inc., 618 F.3d 1025, 1037 (9th Cir. 2010) (survey admissible under Rule 702 raised triable issue of fact). 18 D. The Districts' claims for future damages are legally cognizable. 19 1. The Districts proffer relevant evidence supporting a claim for future damages. 20 21 In each bellwether jurisdiction, a plaintiff may obtain an award of future damages if it proves 22 ongoing injury and the costs of remediation. See, e.g., May v. Holzknecht, 320 S.W.3d 123, 128 (Ky. Ct. 23 App. 2010); Haltiwanger v. Barr, 186 S.E.2d 819, 820 (S.C. 1972); Wilder v. Blue Ribbon Taxicab Corp., 24 719 S.E.2d 703, 708 (S.C. Ct. App. 2011); Zakas v. Jackson, 835 S.E.2d 371, 373 (Ga. Ct. App. 2019); 25 DiLeo v. Nugent, 592 A.2d 1126, 1134–35 (Md. Ct. Spec. App. 1991); Colon v. Robinson, 2014 WL 26 7466566 at *6 (N.J. Super. Ct. App. Div. Jan. 6, 2015); Saide v. Stanton, 659 P.2d 35, 38 (Ariz. 1983). 27 Here, the Districts properly seek future damages under their negligence claims by presenting evidence 28 that (1) Defendants' platforms are reasonably certain to cause ongoing injury and (2) the costs of 205 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 218 of 235 1 addressing those harms. See, e.g., Ex. 1000; see also Ex. 1098. That is all the law requires. 2 The Districts have presented expert testimony from Dr. Hoover explaining that the Defendants 3 have caused numerous ongoing harms that are reasonably certain to recur. See, e.g., Ex. 1000 ¶¶ 54–55 4 (describing "frequent and pervasive" social-media interruptions that undermine instruction), ¶ 61 5 (explaining the "growing burden on teachers" to manage behavioral fallout), ¶ 77 (describing diversion of 6 "limited funding and staffing resources" to address social-media impacts); Ex. 1096 ¶ 55 (opining that 7 Charleston County requires additional resources to address social-media harms); accord; Ex. 1091 ¶ 65; 8 Ex. 1092 ¶ 55, Ex. 1093 ¶ 56; Ex. 1094 ¶ 59; Ex. 1095 ¶ 66; see also Ex. 1000 ¶¶ 30–84. 9 Dr. Hoover next opines that her detailed, 15-year strategic plans are necessary to prevent these 10 future injuries. See Ex. 1091 ¶¶ 106, 110; Ex. 1093 ¶¶ 97, 101; Ex. 1094 ¶¶ 100, 104; Ex. 1095 ¶¶107, 11 111Dr. Leslie then quantifies the costs necessary for each District to implement those plans. See Ex. 609, 12 Ex. 610, Ex. 611, Ex. 612, Ex. 613, Ex. 614. This evidence alone establishes a genuine issue of material 13 fact on future damages. See Haltiwanger, 186 S.E.2d at 820. 14 Defendants cite no authority requiring plaintiffs to prove they can fund mitigation measures absent 15 a damages award. No such rule exists, and for good reason: It would be illogical for Defendants to be 16 absolved just because the entities they harm are underfunded public institutions. Defendants cite three 17 groups of cases, each less relevant than the last. 18 First, they cite cases in which damages were denied because the asserted future injury was 19 speculative, not reasonably certain. See Mauro v. Raymark Indus., Inc., 561 A.2d 257 (N.J. 1989); 20 Battaglia v. United Parcel Serv., Inc., 70 A.3d 602 (N.J. 2013); J.N. Legacy Grp., Inc. v. City of Dallas, 21 745 S.E.2d 721, 729 (Ga. Ct. App. 2013); Pierce v. Johns-Manville Sales Corp., 464 A.2d 1020, 1026 22 (Md. 1983); Davidson v. Miller, 344 A.2d 422, 427 (Md. 1975). These cases are inapposite, given the 23 Districts have presented evidence indicating their future injuries are reasonably certain. See, e.g., Ex. 1000 24 ¶ 8 ("student social media use has had—and continues to have—a profound and detrimental impact"); Ex. 25 1099 ¶¶ 72, 135, 175. At a minimum, whether a future injury is reasonably certain is a question for the 26 jury. See Mauro, 116 N.J. at 139. 27 Second, pertinent to Tucson's case, Defendants cite Arizona cases involving lost profits—Gilmore 28 v. Cohen, 386 P.2d 81 (Ariz. 1963); McAlister v. Loeb & Loeb, LLP, 571 P.3d 891 (Ariz. 2025); Rancho 206 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 219 of 235 1 Pescado, Inc. v. Nw. Mut. Life Ins. Co., 680 P.2d 1235 (Ariz. Ct. App. 1984); and McNutt Oil & Refin. 2 Co. v. D'Ascoli, 281 P.2d 966, 970-971 (Ariz. 1955). But as those cases explain, a plaintiff needs "such 3 precision as, from the nature of the claim and available evidence, is possible." Gilmore, 386 P.2d at 83. 4 And lost profits are distinct because they are amenable to "mathematical precision"—they may be proven 5 through "books of account," "informal memoranda of previous transactions," or "past income tax returns." 6 Gilmore, 386 P.2d at 82–83. The Districts do not seek lost profits and, in any event, Dr. Leslie has provided 7 precise calculations of the costs necessary to address the ongoing harms caused by Defendants' platforms. 8 See Ex. 614 ¶¶ 19-20. 9 Third, Defendants cite cases that simply do not involve future damages at all, and hence are 10 entirely irrelevant to the question at hand. See Nobles v. Jiffy Mkt. Food Store Corp., 579 S.E.2d 63 (Ga. 11 Ct. App. 2003) (concerning liquidated damages award, not future losses); see also J&W Corp. v. Broad 12 Creek Marina, LLC, 896 S.E.2d 328 (S.C. Ct. App. 2023); Accessory Overhaul Grp., Inc. v. Mesa Airlines, 13 Inc., 994 F. Supp. 2d 1296 (N.D. Ga. 2014); Carter v. Wallace & Gale Asbestos Settlement Tr., 96 A.3d 14 147, 157 (2014); Coury Bros. Ranches, Inc. v. Ellsworth, 446 P.2d 458 (Ariz. 1968); Hirsh v. Manley, 300 15 P.2d 588 (Ariz. 1956). 16 Here, the evidence establishes ongoing harm with reasonable certainty. Dr. Hoover's plans were 17 specifically designed to address the continuing, systemic injuries Defendants' platforms cause to school 18 operations, learning, and student well-being. See Ex. 1000 ¶ 5. Specifically, Dr. Hoover determined that 19 Defendants' platforms continue to damage the Districts by negatively affecting: 20 • School operations ("reshaped the educational landscape in ways that extend well beyond individual student behavior," Ex. 1000 ¶ 83); 21 • School climate and environment ("pervasive negative impact on schools and the school 22 environment," id. ¶ 34); 23 • Learning and performance ("many teachers report that social media use interferes with their teaching and students' learning," id. ¶ 45; "growing frustration with managing screen use in class," 24 id. ¶¶ 60, 80, 82); 25 • Teaching effectiveness and classroom dynamics ("the strain Defendants' platforms place on teacher-student relationships," id. ¶ 53); 26 • Teacher morale and job satisfaction ("teachers often experience a decline in morale when faced 27 with the constant distractions caused by students' social media use," id. ¶ 58); and 28 207 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 220 of 235 1 • Student mental health ("[s]chools are increasingly burdened with managing the mental health consequences of social media overuse," id. ¶ 78; see id. ¶¶30–82; Exs. 1091–1096 ¶ 4 (each 2 describing how districts have been forced to "expend and redirect already limited resources" to address these harms)). 3 Dr. Hoover concludes these harms will persist and intensify absent substantial intervention, 4 warning that without implementation of her plans, the Districts will "perpetuat[e] ineffective approaches" 5 and remain unable "to coordinate services, evaluate impact, and make data-informed decisions." See Ex. 6 1000 ¶¶ 8–9, 125; see also Ex. 1102 ¶¶ 35, 46, 60; Ex. 1103 ¶¶ 32, 51; Ex. 1101 ¶¶ 33, 58; Ex. 1104 ¶ 44; 7 Ex. 1105 ¶¶ 39, 137; Ex. 1100 ¶¶ 56, 129. She emphasizes that her 15-year, multi-tiered framework is 8 necessary to prevent and mitigate escalating harms within school environments. Combined with Dr. 9 Leslie's detailed cost analysis, this evidence provides a concrete, non-speculative basis for a future- 10 damages award. 11 Defendants' assertion that Dr. Hoover's plans are merely optional "recommendations" 12 mischaracterizes her work (and diminishes what is, at a minimum, a genuine dispute of fact for the jury 13 to resolve). Dr. Hoover makes clear that the measures she proposes are evidence-based requirements, not 14 flexible suggestions. As she explains, "the staffing and cost recommendations [she] present[s] are not 15 aspirational targets or flexible guidelines; they represent the level of investment and capacity required to 16 adequately address the harms due to social media use in school communities." Ex. 1101 ¶ 41. Across her 17 reports, Dr. Hoover underscores that these recommendations are grounded in empirical data, 18 implementation science, and public-health precedent, reflecting the actual scale of intervention needed to 19 remediate the harms caused by Defendants' platforms. See also Ex. 1100 ¶ 189 ("not aspirational" but 20 "grounded in evidence of what is needed to meet the behavioral and mental-health needs of students"); 21 Ex. 1102 ¶ 45 (providing "evidence-informed recommendations for the staffing, training, and system 22 design needed to meet" the harms); Ex. 1105 ¶ 32 (recommendations are "realistic and proportionate to 23 respond" to social-media harms); Ex. 1103 ¶ 41; Ex. 1104 ¶ 32. 24 Nor is it surprising that the Districts have not yet implemented Dr. Hoover's plans. These are 25 resource-constrained public school systems seeking damages precisely because they cannot fund such 26 measures without relief. Dr. Hoover's role was not to design a plan limited by current budgets but to 27 identify what is necessary to remediate the harms Defendants have caused. See, e.g., Ex. 1105 ¶ 27 ("My 28 208 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 221 of 235 1 role is not to budget based on current constraints, but to provide an evidence-informed benchmark for the 2 staffing levels necessary to meet the scale and specificity of social media-related student distress."). That 3 other districts have not yet implemented identical frameworks is immaterial (and, in any event, may just 4 demonstrate their comparable budget constraints); Dr. Hoover's plans are tailored to these Plaintiffs' 5 unique needs and the specific harms caused by Defendants' platforms. The fact that no other district has 6 had the funding or occasion to request such plans underscores their necessity, not their novelty. In any 7 event, Defendants' argument ignores the over one thousand other school districts who, like the Districts, 8 have brought suit against Defendants seeking comparable relief. 9 Defendants' claim that Districts cannot hire sufficient personnel is likewise baseless. Dr. Hoover 10 directly refutes it: "[C]ontrary to claims that the proposed staffing levels are unattainable, current data 11 reflect a steady flow of qualified school mental-health professionals entering the field." Ex. 1099 ¶ 48; 12 see also Ex. 1102 ¶ 41; accord Exs. 1100 ¶ 40; 1101 ¶ 38; 1103 ¶ 34; 1104 ¶ 34; 1105 ¶ 31. 13 Finally, Defendants' argument that Tucson and Charleston have not yet budgeted for social-media- 14 specific interventions is irrelevant. As Dr. Hoover explains, these Districts "lack the resources and 15 infrastructure to address the modern, external threat to student well-being and school functioning" within 16 their standard budgets. Ex. 1100 ¶ 175; see also id. ¶ 182 ("that social-media harms are not prominently 17 featured in older internal documents or district budgets further supports my conclusion that districts have 18 not had the systems, staffing, or awareness to adequately monitor or address these harms"); Ex. 1101 ¶ 19 128 ("the fact that a district has not fully incorporated a new and emerging harm into its formal systems 20 ... does not mean the harm does not exist or is not substantial"). Similarly, Irvington's historic 21 understaffing has only been exacerbated by Defendants' misconduct. See Ex. 1105 ¶¶ 26–28. 22 In sum, the Districts have presented robust, expert-supported evidence of ongoing, reasonably 23 certain future harm, and the precise costs necessary to remedy them. These are issues of fact for a jury, 24 not matters for summary judgment. Fuller v. Idaho Dep't of Corr., 865 F.3d 1154, 1161 (9th Cir. 2017). 25 26 27 28 209 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 222 of 235 1 2. Breathitt, DeKalb, Harford, and Tucson's nuisance claims permit the abatement they seek. 2 a) Abatement is a distinct equitable remedy designed to eliminate 3 ongoing and future public harms. 4 Breathitt, DeKalb, Harford, and Tucson seek the equitable remedy of abatement to address 5 ongoing and future harms caused by Defendants' nuisance. See Restatement (Second) of Torts § 821C(2) 6 (1979). In addition to asserting negligence claims, Breathitt, DeKalb, Harford, and Tucson bring claims 7 for public nuisance, which rest on a different legal theory. Negligence requires proof that Defendants 8 owed and breached a duty. See MTD Order, 756 F. Supp. 3d at 972. Remedies are limited to money 9 damages (any may include future damages, see supra). Nuisance requires proof that Defendants created 10 "an unreasonable interference with a right common to the general public," and, if proven, authorizes the 11 Court to fashion an equitable remedy. See Restatement (Second) of Torts §§ 821B–C (1979). 12 Abatement is a court-crafted equitable remedy, wholly distinct from damages. City of Huntington, 13 W. Virginia v. AmerisourceBergen Drug Corp., 2025 WL 3009526, at *12 (4th Cir. Oct. 28, 2025) 14 ("[P]ublic nuisance claims serve [the] function [of] focusing on harm suffered by the public more 15 generally and seeking abatement of that harm through the court's equitable authority."); State of Wash. v. 16 McKesson Corp. et al., No. 19-2-06975-9 (Sup. Ct. Wash. July 6, 2021) (granting motion to strike jury 17 demand for abatement because it is "equitable in nature, and should be tried to the Court"). As explained 18 in In re Nat'l Prescription Opiate Litig., "[t]he goal [of abatement] is not to compensate the harmed party 19 for harms already caused by the nuisance. This would be an award of damages. Instead, an abatement 20 remedy is intended to compensate the plaintiff for the costs of rectifying the nuisance, going forward." 21 2019 WL 4043938, at *1 (N.D. Ohio Aug. 26, 2019). Because abatement may "be more certain, prompt, 22 or efficient than the monetary damages [they] seek[], but may ultimately not attain," it is proper for the 23 Districts to pursue both damages and abatement. See Coleman v. Mondelez Int'l Inc., 554 F. Supp. 3d 24 1055, 1065 (C.D. Cal. 2021) (plaintiffs may pursue equitable relief and damages under separate claims). 25 Here, the Districts ask the Court to create an abatement plan to eliminate the nuisance in their 26 schools. Guided by the testimony of Dr. Hoover and Dr. Leslie, this plan would be a post-trial, Court- 27 crafted remedy akin to an injunction. See id. Unlike a damages award, abatement allows the Court to order 28 and oversee implementation to address Defendants' ongoing harms. See Andino v. Apple, Inc., 2021 WL 210 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 223 of 235 1 1549667, at *5 (E.D. Cal. Apr. 20, 2021) ("Money damages are an inadequate remedy for future harm."). 2 Defendants argue that potential recovery of future damages bars abatement. They rely on Sonner 3 v. Premier Nutrition Corp. but misread its holding. Sonner merely provides that a plaintiff cannot "secure 4 equitable" relief if an adequate remedy at law exists. 971 F.3d 834, 844 (9th Cir. 2020) (emphasis added). 5 At trial, a plaintiff may pursue both (as cases interpreting Sonner make clear). Cepelak v. HP Inc., 2021 6 WL 5298022, at *2 (N.D. Cal. Nov. 15, 2021) (allowing damages and equitable relief in the alternative); 7 Coleman, 554 F. Supp. 3d at 1065; Ostrovskaya v. St. John Knits, Inc., 2022 WL 2102895, at *5 (C.D. 8 Cal. Mar. 31, 2022); Roper v. Big Heart Pet Brands, Inc., 510 F. Supp. 3d 903, 917 (E.D. Cal. 2020). If a 9 jury ultimately awards future damages duplicative of the abatement sought, the Court may decline to 10 impose an additional equitable remedy. In various cases cited by Defendants, that was the case: The 11 equitable relief sought was wholly duplicative of the damages requested. See Wood v. Marathon Refining 12 Logistics Serv. LLC, 2024 WL 2242688, at *10 (N.D. Cal. Mar. 21, 2024) (barring plaintiff from seeking 13 equitable restitution that merely duplicated past damages); Rhynes v. Stryker Corp., 2011 WL 2149095, 14 at *4 (N.D. Cal. May 31, 2011) (plaintiff failed to show damages would not provide full compensation).17 15 By contrast, the abatement the Districts seek uniquely addresses ongoing and future harms by allowing 16 the Court to fashion a remedy that eliminates the nuisance. 17 In any event, the determination of whether abatement is or is not duplicative of a damages award 18 is properly made post-trial. See Teutscher v. Woodson, 835 F.3d 936, 957 (9th Cir. 2016) (holding that a 19 plaintiff may seek backward-looking damages from a jury and a "forward-looking remedy . . . from the 20 court in equity"). Accordingly, Breathitt, DeKalb, Harford, and Tucson may pursue equitable abatement 21 through their public nuisance claims. 22 b) Dr. Hoover's strategic plans provide the framework for crafting an abatement remedy. 23 As explained in the Districts' opposition to Defendants' motion seeking to exclude Dr. Hoover 24 and other experts ("SD Daubert Opp."), Dr. Hoover's plans were developed consistently with her standard 25 26 17 Defendants also cite two entirely inapposite cases in support of their argument—Robinson v. C.R. Bard, Inc., 2016 WL 3361825 (N.D. Cal. June 17, 2016) (where the plaintiff had not pursued any timely claims) 27 and Mitsubishi Int'l Corp. v. Cardinal Textile Sales, Inc., 14 F.3d 1507, 1518 (11th Cir. 1994) (concerning 28 the suis generis procedural and equitable context of a pre-judgment constructive trust on assets). 211 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 224 of 235 1 practice and are grounded in public health and implementation science. They outline in specific detail the 2 staffing and programs needed to prevent and mitigate the harms caused by Defendants—providing 3 comprehensive, district-wide interventions, including universal policies for all students, staff, and 4 families, and more intensive programs for those with greater needs. This approach is essential because the 5 harms from Defendants' platforms "pervade the school[s]" and require proactive services "for those 6 students who may not have yet used a platform directly themselves." Ex. 1097B at 623:10–624:4; see also 7 Ex. 1000 ¶ 34. Dr. Hoover further opines that a 15-year timeline is the "necessary duration for 8 implementing a comprehensive, district-led strategic plan to prevent and mitigate the negative impact of 9 social media on school districts, schools, the school environment, student well-being and learning," given 10 comparable public health initiatives, the developmental K–12 timeframe, and system change 11 requirements. Ex. 1000 ¶ 7; see also ¶ 33 (describing a 15-year integrated mitigation plan). 12 Dr. Hoover's strategic plans provide the Court with detailed guidance for crafting an abatement 13 remedy for Breathitt, DeKalb, Harford, and Tucson. See Ex. 1091 ¶¶ 106, 110; Ex. 1093 ¶¶ 97, 101; Ex. 14 1094 ¶¶ 100, 104; Ex. 1095 ¶¶ 107, 111; see also In re Nat'l Prescription Opiate Litig., 2019 WL 4043938, 15 at *1 (N.D. Ohio Aug. 26, 2019) (noting that abatement is an equitable remedy fashioned by the Court 16 using expert testimony and trial evidence). Further, Dr. Leslie calculated the cost to implement Dr. 17 Hoover's plan for each District. This evidence should also assist the Court in crafting an abatement 18 remedy, as monetary funds are an appropriate form of abatement. See infra. In short, the Districts have 19 offered precisely what is required—expert guidance to assist the Court in crafting an abatement remedy 20 (1) The abatement remedy may include monetary relief. 21 Defendants wrongly argue that abatement cannot include monetary relief. Courts have long 22 rejected the "myth" that equitable remedies never involve payment: "the assumption that 'equitable 23 remedies are always orders to act or not to act, rather than to pay, is a myth; equity often orders payment.'" 24 United States v. Apex Oil Co., 579 F.3d 734, 736 (7th Cir. 2009). 25 As an equitable remedy, abatement may be "mold[ed] to the circumstances of [the] particular 26 case." Weinberger v. Romero-Barcelo, 456 U.S. 305, 312 (1982). Courts addressing public health crises 27 have recognized that abatement may include a monetary fund. In the opioid litigation, the Fourth Circuit 28 held that "the remedy of abatement . . . permit[s] a monetary award to fund . . . abatement efforts," 212 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 225 of 235 1 emphasizing that "[c]ourts have broad powers to effect equitable relief." City of Huntington, W. Virginia 2 v. AmerisourceBergen Drug Corp., 2025 WL 3009526, at *20 (4th Cir. Oct. 28, 2025). Likewise, in In re 3 JUUL Labs, Inc., Mktg., Sales Pracs., & Prods. Liab. Litig., the court found that a nuisance abatement 4 remedy may "require defendants to expend the money necessary to abate the nuisance" and ordered 5 contributions to an abatement fund to address youth vaping. 497 F. Supp. 3d 552, 653 (N.D. Cal. 2020). 6 That is precisely what the Districts seek here—a post-trial, Court-crafted remedy guided by Dr. Hoover's 7 and Dr. Leslie's testimony regarding the measures and costs necessary to abate the nuisance. Defendants' 8 effort to distinguish JUUL fails; while the JUUL plaintiffs proposed only an outline, the Districts have 9 presented detailed, expert-supported abatement plans for the Court's consideration. 10 Recent decisions confirm the same principle. In Mayor and City of Baltimore v. Purdue Pharma, 11 L.P., No. 24-C-18-000515, slip op. at 1, 23 (Md. Cir. Ct. Aug. 8, 2025), the court crafted a post-trial 12 abatement plan funded by monetary payments to hire staff and expand treatment and harm-reduction 13 services. Similarly, Dr. Hoover's plans propose staffing and training programs to prevent and mitigate 14 harms caused by Defendants' platforms. See, e.g., Ex. 1000 ¶¶ 6, 83–84, 95–122; Ex. 1093 ¶¶ 74–85. 15 Defendants attempt to distinguish the Purdue Pharma plan by noting that the court did not fund every 16 proposed treatment and service cost. This misses the point. The Purdue Pharma court conducted a fact- 17 based, post-trial inquiry, guided by the jury's findings, to determine which treatments and services should 18 be included—and it did incorporate multiple treatment and recovery programs into its plan. This Court 19 should engage in a similar post-trial evaluation to determine the appropriate scope of the abatement plan. 20 In re Nat'l Prescription Opiate Litig., 622 F. Supp. 3d 584, 606 (N.D. Ohio 2022), rev'd on other 21 grounds, 2025 WL 354758 (6th Cir. Jan. 31, 2025), likewise held that abatement may require payment, 22 explaining that "eliminating a hazard that continues to cause prospective harm . . . will, in virtually all 23 cases, cost a liable defendant some amount of money," and that an equitable abatement award "forces a 24 liable defendant to clean up the mess it made." Id. Again, that is exactly what the Districts seek. While 25 Defendants disingenuously assert that In re Nat'l Prescription Opiate Litig. was vacated, no appellate 26 court has overturned any of the court's rulings on abatement. The appeal addressed only whether an Ohio 27 state statute barred the underlying public nuisance claim, an issue wholly irrelevant here. 28 213 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 226 of 235 1 Other courts agree that monetary abatement remains equitable and confirm that this Court may— 2 and should—fashion a monetary abatement remedy tailored to these harms. See State of Wash. v. 3 McKesson Corp., No. 19-2-06975-9, slip op. at 2 (Wash. Super. Ct. July 6, 2021) (quoting State ex rel. 4 Dep't of Ecology v. Anderson, 620 P.2d 76 (Wash. 1980)) ("a claim remains equitable even though a 5 money judgment might form a part of the relief asked"); People v. ConAgra Grocery Prods. Co., 227 Cal. 6 Rptr. 3d 499 (Ct. App. 2017) (requiring defendants to fund an abatement account to carry out remediation). 7 (2) Defendants' state law authorities are inapposite. 8 Defendants' reliance on state cases to limit this Court's authority is misplaced. Their own 9 authority, Sonner, confirms that "state law cannot expand or limit a federal court's equitable authority," 10 even in diversity cases. 971 F.3d at 841. Federal courts possess "broad powers to tailor and fashion 11 appropriate equitable remedies." United States v. Price, 688 F.2d 204, 211 (3d Cir. 1982); Guthrie v. 12 Transamerica Life Ins. Co., 561 F. Supp. 3d 869, 873 (N.D. Cal. 2021). 13 Even if state law were relevant, state courts have recognized that abatement can extend beyond 14 simply stopping or removing the nuisance. For example, Brown v. City of Phoenix, 557 P.3d 321, 327 15 (Ariz. Ct. App. 2024) distinguishes between the removal and the abatement of nuisances and dealt with 16 an injunction requiring both the removal of certain individuals and the cleanup of hazardous materials 17 they left behind. City of Safford v. Seale, 2009 WL 3390172, at *2 (Ariz. Ct. App. Oct. 21, 2009) also 18 recognizes that abatement can include remediation as well as removal. Similarly, while Spaw, LLC v. City 19 of Annapolis, 156 A.3d 906, 931 (Md. Ct. App. 2017) did not consider the proper scope of abatement 20 remedies, it recognized in dicta that requiring the defendant to not just halt the offending conduct but 21 affirmatively undertake action to undo the effects of the violation would be an appropriate form of 22 abatement. See also Adams v. Comm'rs of Town of Trappe, 102 A.2d 830 (Md. Ct. App. 1954) (requiring 23 defendant to not just stop sale of gasoline but remove tank and pump). 24 Additionally, none of the cases cited by Defendants held that enjoining conduct is the only 25 acceptable form of abatement. In Superior Farm Mgmt., L.L.C. v. Montgomery, 513 S.E.2d 215, 217-18 26 (Ga. 1999), the nuisance was planned but had not yet begun, so no other form of abatement other than an 27 injunction was possible. See also Brandes v. Mitterling, 196 P.2d 464 (Ariz. 1948) (plaintiffs sought only 28 injunction against future use); Cactus Corp. v. State ex rel. Murphy, 480 P.2d 375 (Ariz. Ct. App. 1971) 214 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 227 of 235 1 (same, and nuisance was present for less than two weeks). Meanwhile, Martin v. Howard Cnty., 709 A.2d 2 125 (Md. 1998) and J.D. Jewell, Inc. v. Hancock, 175 S.E.2d 847 (Ga. 1970), address, respectively, the 3 difference between legal and equitable remedies and jurisdiction and venue, not the appropriate forms of 4 abatement. Likewise, Green Meadows Hous. Partners, LP v. Macon-Bibb Cnty., 906 S.E.2d 430, 441 5 (2024) turned on the statutory scope of receiverships, not the breadth of available abatement remedies. 6 (3) Defendants' attacks on Dr. Hoover's plans lack merit. 7 Defendants wrongly claim that Dr. Hoover's plans are untethered to Defendants' conduct because 8 they include comprehensive mental health services. That argument ignores national school mental health 9 standards and implementation science, both of which require a comprehensive approach. See Ex. 1099 ¶ 10 3 (multi-tiered system supports ("MTSS") are "tailored" to meet the "distinct behavioral, emotional, and 11 cognitive risks introduced by social media and the harm to school districts and the school environment"); 12 id. ¶ 67 (MTSS are supported by "public health implementation science, and evidence-informed school 13 mental health frameworks" and "peer-reviewed studies"). Moreover, Dr. Hoover directly links her plans 14 to Defendants' misconduct. She explains: "Educators report struggling to maintain engagement and 15 classroom cohesion amid pervasive digital distractions, while students themselves are contending with 16 anxiety, depression, negative body image, and sleep deprivation—all caused or exacerbated by social 17 media use. These individual-level challenges converge to strain school-wide resources, forcing districts 18 to divert limited staffing and funding to address the mental health and behavioral consequences of digital 19 overuse." Ex. 1000 ¶ 83; see also Ex. 1091 ¶ 4; Ex. 1092 ¶ 4; Ex. 1093 ¶ 4; Ex. 1095 ¶ 4; Ex. 1096 ¶ 4. 20 Defendants' critique of the plans' 15-year duration fares no better. The plans target harm to the 21 Districts, not individual students, and are designed to prevent and mitigate systemic harm over time. See, 22 e.g., Ex. 1000 ¶ 6. Implementation science supports the 15-year timeline as the "necessary duration for 23 implementing a comprehensive, district-led strategic plan . . . [allowing] for the full cycle of planning, 24 implementation, evaluation, and sustainability." Ex. 1000 ¶ 7; see Ex. 1099 ¶ 74. As Dr. Hoover explains, 25 the harms "pervade the school," and "it [is] essential to put services in place for those students who may 26 not have yet used a platform directly themselves." Ex. 1097B at 623:10–624:4; id. at 627:2–628:7. 27 Defendants' "wish-list" characterization is equally baseless. Dr. Hoover makes clear that her 28 staffing recommendations respond to new and exacerbated harms created by Defendants' platforms, not 215 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 228 of 235 1 to historic understaffing. Ex. 1103 ¶ 40 (plans include "additional staffing . . . necessary to respond to the 2 mental health and learning impacts of social media"), ¶ 49 (Defendants' platforms "have not replaced 3 historical challenges; they have created new challenges and have magnified previous challenges"). Her 4 recommendations are a separate MTSS framework targeted specifically to social-media-related harms. 5 See, e.g., Ex. 1093 ¶ 7 ("Existing staffing and programming . . . are insufficient to address the complex 6 impacts of social media. A comprehensive and sustainable plan requires new, dedicated staffing and 7 professional development."), ¶ 27 ("My recommendations are additive, not duplicative . . . targeted toward 8 addressing harms that existing efforts were never designed to address, including those stemming from 9 compulsive platform use."), ¶ 40 (current staff face "role dilution" because of new harms from 10 Defendants' platforms; "the solution is . . . to expand capacity"). 11 Ultimately, Defendants' objections raise factual disputes appropriate for cross-examination, not 12 summary judgment. See Fuller, 865 F.3d at 1161. 13 3. The Districts' harms are not derivative. 14 As this Court has already held, the Districts seek recovery for harms they have themselves suffered 15 and continue to suffer due to Defendants' platforms—not for harms suffered by students. See MTD Order, 16 754 F. Supp. 3d at 968 (rejecting Defendants' derivative-injury argument and distinguishing Ass'n of 17 Washington Pub. Hosp. Dists. v. Philip Morris Inc., 241 F.3d 696, 701 (9th Cir. 2001)). Defendants' 18 contrary claim mischaracterizes Dr. Hoover's strategic plans, which directly address the Districts' own 19 operational and financial injuries. See, e.g., Ex. 1000 ¶ 83 (explaining that student-level challenges 20 "converge to strain school-wide resources, forcing districts to divert limiting staffing and funding to 21 address the mental health and behavioral consequences of digital overuse."). 22 That Dr. Hoover's plans include student mental-health services does not render the Districts' 23 harms derivative. Schools are legally and financially responsible for providing those services, and 24 Defendants' platforms have magnified those burdens. See id. ¶ 5. As Dr. Hoover explains, "mental health 25 challenges triggered and exacerbated by social media use have direct implications for schools, as they 26 affect students' academic performance, social skills, and overall well-being." Id. ¶ 85. Defendants ignore 27 that these are institutional harms: disruptions to school operations, depletion of budgets, and deterioration 28 of the educational environment. See id. ¶¶ 83–85 (describing the "profound and far-reaching impact … on 216 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 229 of 235 1 the school environment" and the "considerable resource strain on schools"). 2 In sum, Dr. Hoover's strategic plans are aimed squarely at remedying harms suffered by the 3 Districts themselves, not vicariously at student injuries. As this Court has already held, the Districts "seek 4 recovery of unique damages" they have directly sustained and continue to sustain as a result of 5 Defendants' misconduct. See MTD Order, 754 F. Supp. 3d at 968. 6 IV. CONCLUSION 7 For the reasons set forth in this omnibus filing, and the accompanying briefs filed by each District, 8 summary judgment should be denied. Genuine issues of material fact exist as to Defendants' negligence 9 and their creation and maintenance of a public nuisance that burdens schools and communities. A jury is 10 entitled to see and hear the evidence of Defendants' misconduct. 11 12 Respectfully submitted, 13 DATED: November 7, 2025 By: /s/ Previn Warren 14 PREVIN WARREN MOTLEY RICE LLC 15 401 9th Street NW Suite 630 Washington DC 20004 16 Telephone: 202-386-9610 pwarren@motleyrice.com 17 18 LEXI J. HAZAM LIEFF CABRASER HEIMANN & 19 BERNSTEIN, LLP 275 BATTERY STREET, 29TH FLOOR 20 SAN FRANCISCO, CA 94111-3339 Telephone: 415-956-1000 21 lhazam@lchb.com 22 Co-Lead Counsel 23 CHRISTOPHER A. SEEGER 24 SEEGER WEISS, LLP 55 CHALLENGER ROAD, 6TH FLOOR 25 RIDGEFIELD PARK, NJ 07660 26 Telephone: 973-639-9100 cseeger@seegerweiss.com 27 Counsel to Co-Lead Counsel and Settlement Counsel 28 217 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 230 of 235 1 JENNIE LEE ANDERSON 2 ANDRUS ANDERSON, LLP 3 155 MONTGOMERY STREET, SUITE 900 SAN FRANCISCO, CA 94104 4 Telephone: 415-986-1400 jennie@andrusanderson.com 5 Liaison Counsel and Ombudsperson 6 7 MATTHEW BERGMAN SOCIAL MEDIA VICTIMS LAW CENTER 8 821 SECOND AVENUE, SUITE 2100 SEATTLE, WA 98104 9 Telephone: 206-741-4862 matt@socialmediavictims.org 10 11 JAMES J. BILSBORROW WEITZ & LUXENBERG, PC 12 700 BROADWAY NEW YORK, NY 10003 13 Telephone: 212-558-5500 jbilsborrow@weitzlux.com 14 15 ELLYN HURD SIMMONS HANLY CONROY, LLC 16 112 MADISON AVE, 7TH FLOOR NEW YORK, NY 10016 17 Telephone: 212-257-8482 ehurd@simmonsfirm.com 18 19 ANDRE MURA GIBBS MURA, A LAW GROUP 20 1111 BROADWAY, SUITE 2100 OAKLAND, CA 94607 21 Telephone: 510-350-9717 22 amm@classlawgroup.com 23 MICHAEL M. WEINKOWITZ LEVIN SEDRAN & BERMAN, LLP 24 510 WALNUT STREET SUITE 500 25 PHILADELPHIA, PA 19106 26 Telephone: 215-592-1500 mweinkowitz@lfsbalw.com 27 28 218 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 231 of 235 1 MELISSA YEATES KESSLER TOPAZ MELTZER & CHECK LLP 2 280 KING OF PRUSSIA ROAD 3 RADNOR, PA 19087 Telephone: 610-667-7706 4 myeates@ktmc.com 5 Plaintiffs' Steering Committee Leadership 6 RON AUSTIN 7 RON AUSTIN LAW 400 MANHATTAN BLVD. 8 HARVEY, LA 70058 Telephone: 504-227–8100 9 raustin@ronaustinlaw.com 10 AELISH M. BAIG 11 ROBBINS GELLER RUDMAN & DOWD LLP 1 MONTGOMERY STREET, #1800 12 SAN FRANCISCO, CA 94104 Telephone: 415-288-4545 13 AelishB@rgrd.com 14 PAIGE BOLDT 15 ANAPOL WEISS 130 N. 18TH STREET, #1600 16 PHILADELPHIA, PA 19103 Telephone: 215-929-8822 17 pboldt@anapolweiss.com 18 THOMAS P. CARTMELL 19 WAGSTAFF & CARTMELL LLP 4740 Grand Avenue, Suite 300 20 Kansas City, MO 64112 Telephone: 816-701-1100 21 tcartmell@wcllp.com 22 FELICIA CRAICK 23 KELLER ROHRBACK LLP 1201 THIRD AVENUE, SUITE 3400 24 SEATTLE, WA 98101 Telephone: 206-623-1900 25 fcraick@kellerrohrback.com 26 SARAH EMERY 27 HENDY JOHNSON VAUGHN EMERY PSC 600 WEST MAIN STREET, SUITE 100 28 LOUISVILLE, KY 40202 219 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 232 of 235 1 Telephone: 859-600-6725 semery@justicestartshere.com 2 3 KIRK GOZA GOZA HONNOLD 4 9500 NALL AVE. #400 OVERLAND PARK, KS 66207 5 Telephone: 913-412-2964 Kgoza @gohonlaw.com 6 7 RONALD E. JOHNSON, JR. HENDY JOHNSON VAUGHN EMERY PSC 8 600 WEST MAIN STREET, SUITE 100 LOUISVILLE, KY 40202 9 Telephone: 859-578-4444 rjohnson@justicestartshere.com 10 11 MATTHEW P. LEGG BROCKSTEDT MANDALAS FEDERICO, LLC 12 2850 QUARRY LAKE DRIVE, SUITE 220 BALTIMORE, MD 21209 13 Telephone: 410-421-7777 mlegg@lawbmf.com 14 15 SIN-TING MARY LIU AYLSTOCK WITKIN KREIS & 16 OVERHOLTZ, PLLC 17 EAST MAIN STREET, SUITE 200 17 PENSACOLA, FL 32502 Telephone: 510-698-9566 18 mliu@awkolaw.com 19 JAMES MARSH 20 MARSH LAW FIRM PLLC 31 HUDSON YARDS, 11TH FLOOR 21 NEW YORK, NY 10001-2170 22 Telephone: 212-372-3030 jamesmarsh@marshlaw.com 23 JOSEPH H. MELTER 24 KESSLER TOPAZ MELTZER & CHECK LLP 280 KING OF PRUSSIA ROAD 25 RADNOR, PA 19087 26 Telephone: 610-667-7706 jmeltzer@ktmc.com 27 HILLARY NAPPI 28 AWK ATTORNEYS 220 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 233 of 235 1 1133 WESTCHESTER AVE, SUITE N-224 WHITE PLAINS, NY 10604 2 Telephone: 914-468-4840 3 hnappi@awk-saa.com 4 EMMIE PAULOS LEVIN PAPANTONIO RAFFERTY 5 316 SOUTH BAYLEN STREET, SUITE 600 PENSACOLA, FL 32502 6 Telephone: 850-435-7107 7 epaulos@levinlaw.com 8 RUTH THI RIZKALLA THE CARLSON LAW FIRM, PC 9 1500 ROSECRANS AVE., STE. 500 MANHATTAN BEACH, CA 90266 10 Telephone: 415-308-1915 11 rrizkalla@carlsonattorneys.com 12 ROLAND TELLIS DAVID FERNANDES 13 BARON & BUDD, P.C. 15910 Ventura Boulevard, Suite 1600 14 Encino, CA 91436 15 Telephone: 818-839-2333 rtellis@baronbudd.com 16 dfernandes@baronbudd.com 17 DIANDRA "FU" DEBROSSE ZIMMERMANN DICELLO LEVITT 18 505 20th St North 19 Suite 1500 Birmingham, Alabama 35203 20 Telephone: 205-855-5700 fu@dicellolevitt.com 21 22 Plaintiffs' Steering Committee Membership 23 JOSEPH VANZANDT BEASLEY ALLEN 24 234 COMMERCE STREET MONTGOMERY, AL 36103 25 Telephone: 334-269-2343 26 joseph.vanzandt@beasleyallen.com 27 Federal/State Liaison 28 221 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 234 of 235 1 Attorneys for Individual Plaintiffs 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 222 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR Case 4:22-md-03047-YGR Document 2480 Filed 11/21/25 Page 235 of 235 1 ATTESTATION PURSUANT TO CASE MANAGEMENT ORDER NO. 27 2 I, Previn Warren, attest that the evidence cited herein fairly and accurately supports the facts as 3 asserted. 4 5 Dated: November 7, 2025 By: /s/ Previn Warren 6 7 FILER'S ATTESTATION 8 I, Previn Warren, hereby attest, pursuant to N.D. Cal. Civil L.R. 5-1, that the concurrence to the 9 filing of this document has been obtained from each signatory hereto. 10 11 Dated: November 7, 2025 12 By: /s/ Previn Warren 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 223 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT 4:22-md-03047-YGR

Plaintiffs' Corrected Omnibus Opposition to Defendants' Motions for Summary Judgment, In Re: Social Media Adolescent Addiction/Personal Injury Products Liability Litigation

Raw text