Meta details AI tools and 2026 exploitation enforcement

Meta said new AI checks helped it action 33.2 million child-exploitation items in early 2026, mostly before user reports.

Meta said it expanded detection of child sexual exploitation on Facebook and Instagram, including LLM checks for ads that covertly signpost off-platform destinations, review or blocking of those destinations, additional AI sweeps, a red-teaming agent, and recidivism detection for banned users. From January to June 2026 it actioned 33.2 million pieces, more than 97% before any user report. In India it actioned 5.3 million pieces and said cases would be reported to the National Cyber Crime Reporting Portal alongside existing NCMEC reporting. TechCrunch added that more than 98% of the India items were found proactively, a figure Meta’s newsroom summary did not give. Only TechCrunch reported that in August Meta agreed to pay up to $18 billion to settle a child-safety lawsuit with 29 U.S. states.

• Meta actioned 33.2 million child sexual exploitation items on Facebook and Instagram from January to June 2026.
• More than 97% of those items were found proactively, before any user report.
• New LLM checks target covert ad “signposting” that routes users off-platform; Meta also reviews or blocks ad destinations, not only ad content.
• Meta described further AI sweeps, a red-teaming AI agent that tests its safeguards, and better detection of banned users returning on new accounts.
• In India, Meta actioned 5.3 million pieces; TechCrunch said more than 98% were found before reports, a rate Meta’s newsroom item did not state.
• Meta said Indian cases will be reported to the National Cyber Crime Reporting Portal, alongside existing NCMEC reporting.
• TechCrunch reported that in August Meta agreed to pay up to $18 billion to settle a child-safety lawsuit with 29 U.S. states; Meta’s newsroom item did not mention that deal.