Meta is rolling out AI tools to detect ads that direct users to child sexual abuse material outside its platforms on 7 October, TechCrunch reported. Meta says the tools target a route to illegal content that can begin with an advertisement that looks harmless.
Key points
- Meta says a new large language model system detects ads that point users towards illegal content elsewhere online.
- The company is examining ad destinations as well as the ads themselves.
- Meta says its systems found more than 97% of the 33.2 million pieces of child sexual exploitation content it acted on across Facebook and Instagram in the first half of 2026 before users reported them.
- Meta is establishing a direct route for reporting suspected cases in India.
Meta checks where ads lead
Meta calls the tactic “signposting”: an ad can appear ordinary while pointing people towards illegal content or other harmful activity on another site. The company says it has seen bad actors use this approach to evade detection. An ad need not contain child sexual abuse material itself to lead users to a site hosting it.
Meta says its new large language model system is designed to find those ads. It is also examining the destinations behind them, giving the company a way to block websites or other destinations that break its rules and act against the accounts responsible. That moves the check beyond the material visible in an ad.
The company says additional AI scans are intended to find child exploitation content its earlier systems may have missed. Meta is also improving how it identifies people who open new accounts after their previous accounts have been removed.
Another tool, which Meta calls a “red-teaming AI agent”, tests its safety measures for weaknesses that bad actors might exploit. Meta says the testing could help it identify methods of abuse before they become more common.
Meta reports 33.2 million content actions
TechCrunch reported that Meta took action against 33.2 million items of child sexual exploitation material across Facebook and Instagram from January to June 2026, with its systems identifying more than 97% ahead of user reports. Those figures concern content on the platforms; the newly introduced ad checks address routes to material elsewhere online.
In India, Meta acted on 5.3 million pieces during the same period, with more than 98% detected before users reported them. The figures sit alongside continuing scrutiny of child safety on Meta’s services, including lawsuits and criticism from lawmakers.
In August, Meta agreed on a settlement payment of up to $18 billion for a child safety lawsuit involving 29 U.S. states, TechCrunch reported. The company has also introduced parental controls for Meta AI and pre-teen accounts on WhatsApp this year.
India gains a direct reporting route
Meta is establishing direct reporting integrations with the Indian Cybercrime Coordination Centre under the Ministry of Home Affairs, NDTV Profit reported. Suspected child safety and child sexual abuse material cases originating in India are to be routed through the National Cyber Crime Reporting Portal.
