Meta & Children's Online SafetyPLAINTIFFS’ CHARACTERIZATION
2014–2026
Evidence timeline
539 passages quoted from Meta's internal records, court filings and congressional testimony, in date order. Each is labelled with its source.
Internal documents only Youth Targeting Concealment Knowledge Knowledge Concealment Youth Targeting Knowledge Knowledge Youth Targeting Profits over Safety Profits over Safety Knowledge Knowledge Profits over Safety Profits over Safety Youth Targeting Profits over Safety Knowledge Profits over Safety Youth Targeting Knowledge Concealment Knowledge Knowledge Knowledge
2021
PLAINTIFFS’ CHARACTERIZATION
In 2021, Meta contracted with a vendor to conduct a survey of preteens. Meta instructed the vendor to not inform Meta employees if any of the preteen survey subjects were on Instagram, so that Meta "as a company won't be made aware of under 13." …
PLAINTIFFS’ CHARACTERIZATION
During 2021, 6.7% of surveyed Instagram users had seen self-harm content within the last seven days. For users between 13 and 15 years of age, 8.4% had seen content relating to self-harm on Instagram within the last seven days.
Suicide & self-injuryContent Algorithm
PLAINTIFFS’ CHARACTERIZATION
In an internal Meta survey in or around 2021, 6% of teen girls in the U.S. and 13% of teen girls in the U.K. "traced their desire to self-harm/commit suicide to Instagram."
Suicide & self-injury
PLAINTIFFS’ CHARACTERIZATION
In an internal chat from 2021, Meta employees suggest censoring terms such as "mental health" in internal documents that "might become public" to avoid scrutiny.
"Teens are the target audience for Reels, but we have limited knowledge, tech, and tools to keep them safe."
Child sexual exploitation
"TLDR: The Reels ecosystem makes it more vulnerable to integrity challenges" "1 There are multiple Reels teams launching a high volume of product improvements, leading to increases in sexually suggestive and objectionable content." "2 The focus on entertainment / mimicry, and the teen audience leaves Reels more open to 1) sexually suggestive and offensive content and 2) vulnerability of younger people. Reels teams should build user-centered tools (audience controls, recommendation feedback) and preventive mechanisms (i.e., tuning ranking models, protections for minors) to keep users safe." "3 Teens, women, LGBT people, and communities of color are more vulnerable to harm and marginalization on IG. Early research suggests these inequitable patterns are reproduced on Reels." "There are multiple (15+) teams working on Reels launching a high volume of product improvements (100+ on 2021 H1). This product pace and complexity is leading to greater volatility in integrity, especially increases in sexually suggestive content." "We have less sophisticated knowledge + tools for empowering users to address harm" "Subsurfaces on Reels have different integrity challenges. Example: increases in objectionable content promoted by our ranking models result in bigger spikes especially on Reels Cover and In-Feed Unit (RIFU)." "System for identifying harmful content in Reels is harder than other content review systems. Systems for identifying harmful content in short form videos works less well compared to images and text, takes longer, and is harder to judge/less accurate." "Less is known about what, how, and why certain kinds of integrity challenges come with the surface. For example, while the child safety team has done extensive work to improve child safety generally across IG, Reels is not yet a focus. This is concerning, given much of Reels (and TikTok) came from viral dance trends consumed and created by teens and kids." "Ranking models cause integrity regressions by increasing the prevalence of non-recommendable content. Increasing topic diversity increased all objectionable content, and recommending based on P(click), Mimicry, and downranking watermarked media increased the prevalence of sexually suggestive content." "Trends, Entertainment, and Mimicry can lead to harm. Risky trends (i.e., the TikTok WAP trend for teens or trends mocking disabilities) may leave people open to harm, ranging from seeing harmful content to bullying or predatory behaviors. Offensive content (hate speech, violence, bullying) may also be delivered under the guise of "entertainment" at higher rates on Reels." "Unconnected dancing videos featuring young girls raise safety concerns. Studies and internal feedback suggests that people are uncomfortable with seeing and concerned about the safety of unconnected Reels content of tween and young girls dancing and posing." "Teens are the target audience for Reels, but we have limited knowledge, tech, and tools to keep them safe." "Teens don't think deeply about safety risks until something bad happens."
"We expect to fall short of our original H1 goal of eliminating the 2M backlog of underage report cases"
"We expect to fall short of our original H1 goal of eliminating the 2M backlog of underage report cases, due to slower than expected hiring of ops reps and continued increase in appeals (growing to 2.7M at one point in H1). We are investigating more aggressive automation levers to reduce the backlog. Due to the size of the backlog, we have been cautious to add new or improved reporting mechanisms and plan to launch only 1 of the 2 changes in our H1 goals."
"We began experimentation... to reduce IIC... by preventing ;'potential' violating adults from contacting minors in Reels, Comments, Suggested For You, Explore, and Follows."
Child sexual exploitation
"We began experimentation for 7 initiatives to reduce IIC (Inappropriate Interactions with Children) by preventing 'potential' violating adults from contacting minors in Reels, Comments, Suggested For You, Explore, and Follows."
"[W]e must build some sticky new features that are built for Tweens first in order to make the experience truly compelling to Tweens."
Addiction & problematic use
"We believe that in addition to building an age appropriate experience with guardian supervision we must build some sticky new features that are built for Tweens first in order to make the experience truly compelling to Tweens. We hosted a design sprint to explore new experiences that would be compelling to tweens in four themes: Connection, Identity, Play and Exploration."
"We released a high volume (200+ in H1 2021) of product launches, 30+ leading to integrity regressions"
"There are multiple (15+) teams working on Reels launching a high volume of product improvements (200+ on 2021 H1).This product pace and complexity is leading to greater volatility in integrity." "We released a high volume (200+ in H1 2021) of product launches, 30+ leading to integrity regressions." "Gaps on recommended surface experimentation cause unexpected integrity impacts. Example: biases between CEPd and PD can cause disparities between QE pretests and post-launch topline impact." "Reels presents new and unique integrity system challenges." "Subsurfaces on Reels have different integrity challenges. Example: Reels Tab has higher prevalence of regulated goods, Cover of disturbing content, and RIFU of sexually suggestive content." "System for identifying harmful content in Reels is harder than other content review systems. Systems for identifying harmful content in short form videos works less well compared to images and text, takes longer, and is harder to judge/less accurate." "Features of Reels (see next section) create unexpected challenges for integrity models. For example, the fast paced nature of viral trends require faster recurring training for models to keep up with non-recommendable spikes."
"Meaningful Interactions are highly related to harmful content"
Child sexual exploitationHate speechLikes
"Meaningful Interactions are highly related to harmful content. Interactions such as clicks, likes, private shares, and comments are associated with sexually suggestive and hate speech content across IG." "Integrity issues may cost us engagement — people most often report scrolling past, exiting Reels, and exiting Instagram when having a negative experience on Reels rather than using other remediation tools."
"Sexually Suggestive Content featuring Minors (SSCM) is 2-3x higher on Reels than Explore"
Child sexual exploitationHate speechBullying & harassmentContent Algorithm
"Ranking models cause integrity regressions by increasing the prevalence of non-recommendable content. Increasing topic diversity increased all objectionable content, and recommending based on P(click), Mimicry, and downranking watermarked media increased the prevalence of sexually suggestive content." "Trends, Entertainment, and Mimicry can lead to harm. Risky trends (i.e., the TikTok WAP trend for teens or trends mocking disabilities) may leave people open to harm, ranging from seeing harmful content to bullying or predatory behaviors. Offensive content (hate speech, violence, bullying) may also be delivered under the guise of 'entertainment' at higher rates on Reels." "Unconnected dancing videos featuring young girls raise safety concerns. Studies and internal feedback suggests that people are uncomfortable with seeing and concerned about the safety of unconnected Reels content of tween and young girls dancing and posing." "Sexually Suggestive Content featuring Minors (SSCM) is 2-3x higher on Reels than Explore*. Female and unknown gender teens, as well as unknown gender non-teens, create the most SSCM. However, we're also over-enforcing by labeling birth announcements and parenting videos under this category. (DM me for quip on Protecting Youth!)" "The highest perceived reach integrity types are the least detected, and vice versa. People didn't often report the highest labeled prevalence integrity types (sexually suggestive, violence, graphic content). Similarly, hate speech and bullying are some of the lowest detected prevalence on Reels, but participants reported witnessing those more often and report them as being more upsetting." "Female and 'Unknown Gender' Teens create 73% of 'SSCM' content created by teens... ...While male teens report the most. 'Custom Gender' Teens both create (1.2%) and report (1.9%) 'SSCM' content at higher rates than their DAU (.4%)." "Benign but still risky content featuring minors: Children dancing/posing unto itself is not problematic, but it becomes an issue if bad actors see it. Tolerance for this content varies by country and is not preferred for many non-teen users. Preventing IIC actors from seeing/interacting with minors, more effective / child content personalization (SFPLT) or creating 'teen-centered' ranking approaches for teens may help solve this problem."
"The stance we have historically taken is to give people control, but not in a way that hurts metrics."
Addiction & problematic useInfinite ScrollAutoplay
"The stance we have historically taken is to give people control, but not in a way that hurts metrics."
"Review of the account photos to determine if user is <13 years old, and if so, user is checkpointed... Reason for backlog: Minimum investment in u13 in previous quarters"
"Build efficiency and decrease our ID verification backlog from ~500k to <6k jobs through automation (both auto-reject and auto-accept)." "Review of the account photos to determine if user is <13 years old, and if so, user is checkpointed" "Size: ~780k" "Goal: Review jobs by 7days/48hours [TBC] by June" "Reason for backlog: Minimum investment in u13 in previous quarters" "User aiming to change their age within and across the 18 year old threshold (not including U13)" "Mitigation: TBC; investigating inconsistency of ID use in protocols" "Goal: Review jobs by 7days/48hours [TBC] by June" "Reason for backlog: Minimum investment in u13 in previous quarters"
January 2021
"We could reduce precision thresholds for quarantine or auto-delete but... we do not recommend this"
Child sexual exploitation
"We could reduce precision thresholds for quarantine or auto-delete but given maturity of classifiers and harsh penalties for child safety offenses, we do not recommend this." "████ (PM) & I are aligned that this is not a viable path forward ATM because of: 1. high risk of FP 2. lack of appeals staffing (so inability to correct account-level disables) In H1, we're scoping out a roadmapping investment on banking efficiencies for child safety violations that do NOT include law enforcement (LE) reporting, eg minor sexualization & benign child nudity, this will free up FTC WHO capacity, which we will re-invest into CEI review with the goal of identifying & banking more "novel" (unmatched) CEI."
"Elder siblings / First child teens play a pivotal role in bringing the rest of the household (parents and younger siblings) to Instagram."
"Elder siblings / First child teens play a pivotal role in bringing the rest of the household (parents and younger siblings) to Instagram." "Tweens rely on their older sibling to create and set up their account, and seek their guidance on a username, profile photo, which accounts to Follow, how to use and think about each surface, and what/ when to comment."
Hidden like counts did not ship because it "was pretty negative to FB metrics"
Likes
"Pure Daisy did not ship because it was not impactful from the lens that IG was evaluating previously and was pretty negative to FB metrics."
Mimicry is a problem for teens and "we are not equipped integrity wise to address this challenge "
"TL;DR - Roughly ~10% of reels associated to trends could come from underage users (<13) - We plan to scale mimicry by amplifying trends in H121 potentially making the size of the problem bigger - It seems that we are not equipped integrity wise to address this challenge yet Problem - Reels format has a particular appeal for teens but as well underage users - Trends trough mimicry seem to seem to exacerbate the problem by making creation accessible to very young audience - We plan to scale mimicry in H121 by amplifying trend distribution, causing a potential risk to make the problem even bigger. - Based on early internal investigation, we are not equipped to properly size / address the problem"
Max Eulenstein: "No one wakes up thinking they want to maximize the number of times they open Instagram that day. But that's exactly what our product teams are trying to do."
"No one wakes up thinking they want to maximize the number of times they open Instagram that day. But that's exactly what our product teams are trying to do."
February 2021
PLAINTIFFS’ CHARACTERIZATION
For example, in February 2021, noted that Meta was working to recruit "Gen Alpha before they reach teenage years" to Instagram.
"Teen girls perceive Instagram as the 'worst' social media app in terms of body and appearance comparison."
Body image & eating disorders
"Teen girls perceive Instagram as the 'worst' social media app in terms of body and appearance comparison." "Appearance-based comparison is common on Instagram. One-third (33%) of people say they compare their appearances to others' often or always. Nearly half (48%) of teen girls do." "It causes people to feel bad about themselves: One quarter (26%) report often or always seeing posts that make them feel worse about their bodies or appearances. More than one-third (37%) of teen girls do." "One-quarter of people on IG (23%) feel a lot or extreme pressure to look perfect on Instagram, and one-third of teen girls (34%) do." "Appearance-based comparison is worse for teens, women of all ages, and people in globally western countries."
"I have not included younger kids (10-12 yos) in this research… since they aren't supposed to be on IG at all"
████: "For context on age selection: I have not included younger kids (10-12 yos) in this research for a number of reasons: 1) Already high risk to kids, 2) Although there are definitely kids this age on IG, I'm concerned about risks of disclosure since they aren't supposed to be on IG at all" ████: "For ages 10-12, I'm not suggesting we find this age group using IG (and I think policy wouldn't allow us to try due to violation)."
"Personally, I think problematic use is a valuable area to invest understand time in. But I'm not a big believer in having a well-being/integrity product team approach. I think it'll require more fundamental changes to our goals, what type of work they incentive, and therefore how core mechanics work (feed design, ranking, sharing, notifs)."
Addiction & problematic useContent AlgorithmNotifications
Kristin Hendrix: "Quick heads up on the WB stuff - Lori was looped into this by virtue of her collabs w/ central. I'm working through it w/ Pratiti and have only asked Lori to share some prior research we've done at IG that has led us to focus on social comparison, bullying, etc." Kristin Hendrix: "Pratiti's concern is that IG is focusing on WB areas that are not ones that came out of the central WB researchers' survey (which Lori was looped in on)." Kristin Hendrix: "I'm working through it w/ Pratiti, but the bigger issue I'm seeing is that there may have been expectations not communicated with us in IG that we would shift our focus areas to align with whatever came out of that survey. That's the piece I'm keeping Lori out of so she's not having to navigate that." Kristin Hendrix: "In other words, Pratiti is surprised IG continues to work on WB issues that her team's survey didn't find to be as high-pri and is asking us to justify that." Adam Mosseri: "I will admit I'm surprised body image and social comparison are as low as they are for internal leadership, US Policy insiders, and Gen Pop" Max Eulenstein: "We considered a more direct approach for self-care/support, but an important factor was that it seemed like stakeholders and gen pop were holding us more directly accountable for social comparison/body image issues." Max Eulenstein: "Personally, I think problematic use is a valuable area to invest understand time in. But I'm not a big believer in having a well-being/integrity product team approach. I think it'll require more fundamental changes to our goals, what type of work they incentive, and therefore how core mechanics work (feed design, ranking, sharing, notifs)."
"Backlog of 780k Underage (u13) review jobs (i.e. where people have reported to us that they think an account is u13). Inbound demand continues to be higher than our supply of reviewers. We are also investigating two adjacent age-related flows for Date of Birth edit on FB and IG... Additionally, we are expecting significantly more incoming jobs following a fixed SEV that was preventing FB users reporting other FB users as being potentially underage. This backlog (and the SEV) is concerning given the Youth XFN work - which making plans to communicate a strong public statement about our stance on u18 safety in June/July."
Child sexual exploitation
"[S223016] Mistakenly autoclosing VHS/EHS jobs (e.g. terrorism, suicide, child nudity), due to a logic mapping issue that was put in place in March as part of WFH response, affecting ~10% of all auto-close traffic (~415K jobs/week). Fix has shipped to 100% and we are investigating to see if we can retroactively get to the content. SEV is ongoing, but immediate issue resolved." "Backlog of 780k Underage (u13) review jobs (i.e. where people have reported to us that they think an account is u13). Inbound demand continues to be higher than our supply of reviewers. We are also investigating two adjacent age-related flows for Date of Birth edit on FB and IG to understand current state and whether this is being addressed in the ID Review backlog war room. Additionally, we are expecting significantly more incoming jobs following a fixed SEV that was preventing FB users reporting other FB users as being potentially underage. This backlog (and the SEV) is concerning given the Youth XFN work - which making plans to communicate a strong public statement about our stance on u18 safety in June/July. Team working on a ops ask to clear the backlog, as well as maintain at an acceptable level given the permanent increases in report volume."
"the nielsen study does show causal impact on social comparison 😬"
████: "the nielsen study does show causal impact on social comparison 😬" ████: "I know - I remember that. specifically remembering ████ saying that's the outcome he stands behind."
Sources include internal documents, public filings and press reports. “Alleged by plaintiffs” denotes the plaintiffs’ characterization, not Meta’s own words. Source labels identify provenance, not a finding of fact.