Meta rolled out new AI tools to detect ads that secretly lead to child sexual abuse material. The tools target ads that look normal but redirect users to illegal content. Meta now examines where an ad sends users, not just its content. A new large language model system detects signposting to harmful sites. A red-teaming AI agent tests Meta’s own safety measures for weaknesses. Meta also improved systems to catch repeat offenders with new accounts. The company acted on 33.2 million pieces of child sexual exploitation content in H1 2026. Over 97% of this content was detected before users reported it. In India, Meta acted on 5.3 million pieces, with over 98% detected proactively. The rollout follows a $18 billion settlement over child safety concerns.
What changed
Meta relied on user reports and content-based detection for harmful ads.
What it unlocks
Detecting ads that appear harmless but redirect users to illegal child abuse material.
- 33.2 million pieces of content acted on in H1 2026
- 5.3 million in India
- 97% detected before user reports
- 98% in India detected before user reports
Sources