Meta deploys AI to detect ads directing users to child abuse material

Meta announced on Wednesday that it is implementing new artificial intelligence tools designed to identify advertisements and accounts that may be used to secretly direct users to child sexual abuse material. This initiative comes as the company reported taking action against 33.2 million pieces of child sexual exploitation content across Facebook and Instagram during the first half of 2026.

According to Meta, more than 97% of the content it acted upon was detected by its internal systems before any user reports were filed. In India specifically, Meta addressed 5.3 million instances of child sexual exploitation content in the same period, with over 98% identified proactively.

The new AI tools are specifically targeting a tactic known as 'signposting,' where advertisements may appear innocuous but are suspected of leading users to illegal content or harmful activities hosted on external websites. Meta's updated approach involves examining where an ad directs users, not solely its content, enabling the company to block offending websites and take action against associated accounts.

Meta is also enhancing its AI-driven scans to capture child exploitation content that might have been missed by previous systems and is committed to incorporating new signals as it learns more about the operational methods of these networks. Additionally, a 'red-teaming AI agent' has been introduced to proactively test Meta's safety measures and identify potential vulnerabilities that malicious actors could exploit.

These enhanced safety measures are being introduced amid ongoing pressure on Meta regarding child safety on its platforms, including lawsuits and criticism from lawmakers. In August, Meta agreed to a settlement of up to $18 billion to resolve a child safety lawsuit involving 29 U.S. states.