Meta said on October 7 that it had added artificial intelligence tools to detect ads that covertly direct users to child sexual abuse material hosted on other platforms

Meta announced on October 7 that it deployed new artificial intelligence tools to detect advertisements that covertly direct users to child sexual abuse material hosted on external platforms, a measure introduced after a BBC Eye investigation in July 2026 found Instagram running such paid advertisements in India. The company reported actioning 5.3 million pieces of child sexual exploitation content on Facebook and Instagram in India between January and June 2026, representing 16 percent of the 33.2 million pieces removed globally, but did not disclose figures for CSAM-linked advertisements specifically. Questions remain about when the new ad-review measures went live, whether human moderators now review ads before publication in India, and whether Meta reported CSAM found in India directly to Indian police as required by the Protection of Children from Sexual Offences Act.

Ad-layer detection closes a revenue-driven vector that content moderation alone cannot reach. Offenders who embed coded language, QR codes or contact handles in paid advertisements bypass traditional keyword filters and image hashes because the call to action is abstracted from the abuse material itself. AI online-safety systems trained on behavioral patterns rather than static content signatures identify these indirection tactics by recognizing anomalous purchase flows, audience-targeting combinations that correlate with known CSAM distribution channels, and advertisement creative that mimics legitimate messaging but directs traffic to high-risk domains. When integrated upstream of ad approval workflows, such detection prevents monetized child-exploitation content from ever reaching a user feed. Guardii technology, built to detect threat patterns across messaging platforms, extends the same sequence-analysis principle to advertising pipelines, offering platforms a detection capability that protects the revenue model without enabling abuse within it.