Back to this hearing · All hearings

Published hearing exchange

Social Media’s Impact on Homeland Security

Testimony

Questioner: Alex Padilla

Narrative: ConcealmentHarm: Hate speechDesign feature: Content Algorithm
Q. · Question

Alex Padilla

A. · Response

Respondent not recorded

QuestionFollowing up on Cox's live-hearing testimony that Meta catches 95% of hate speech before it's reported, Padilla asks (a) where that content originates, (b) how many users saw the uncaught 5% before removal, (c) how long it stayed up, (d) Meta's false-positive rate by language, and (e) what internal success metric Meta uses.

ResponseMeta answers (a) and (e) with its prevalence methodology, but says outright "we do not track this information" for (b) — the specific recommendation-count question Padilla pressed on live — declines to give a direct median-duration figure for (c), and for (d) reports Q2 2022 hate-speech appeal data (2.7M appeals, 238K restored, ~9% reversal rate) without a language breakdown or a distinct "false positive rate."