Meta announced on July 12, 2026 that its AI systems have cleared 53 lakh pieces of child sexual abuse material from Facebook and Instagram between January and June. 98 percent of the offending content was caught before it was reported by users, the company said. The move follows a broader push to apply machine‑learning tools to flag illegal material in real time.
New detection layers use large language models and a network of AI sweeps that scan posts, captions and ad feeds for illicit language patterns. Meta also conducts red‑teaming exercises, deliberately testing its algorithms for blind spots and potential misuse, the policy team added.
The platform has upgraded its recidivism engine to flag repeat offenders, automatically removing content and restricting accounts that breach the rules again. Users who repeatedly post CSAM face heightened scrutiny and longer bans, according to Meta.
Meta said it will continue to expand AI‑driven sweeps and reinforce enforcement across its services, aiming to preempt new violations before they reach public view.