Meta launches AI tools to combat child sexual abuse material via hidden ads

Meta announced on Wednesday that its systems identified 33.2 million pieces of child sexual exploitation content on Facebook and Instagram in the first half of 2026, with over 97% detected before user reports. In India, the company acted on 5.3 million such pieces, with more than 98% found by automated systems. The company is rolling out new AI tools to detect ads that may lead users to illegal child abuse material, a tactic increasingly used by bad actors to evade detection. Meta introduced a large language model to identify 'signposting'—ads that appear harmless but direct users to harmful content. The company is also enhancing its systems to track where ads lead users and block harmful destinations. Additionally, Meta is improving its ability to detect users who return to its platforms with new accounts after previous accounts were removed. These measures come amid ongoing legal and regulatory pressure over child safety, including a $18 billion settlement with 29 U.S. states in August. Meta has also introduced parental controls, pre-teen accounts on WhatsApp, and alerts for self-harm searches on Instagram. The company is using a 'red-teaming AI agent' to test its safety measures and identify potential weaknesses. Meta emphasized its commitment to learning more about how these networks operate to refine its strategies. The new tools aim to address evolving tactics by bad actors, ensuring greater protection for users on its platforms.