Meta's automated ad review system approved 332 advertisements containing AI-generated child sexual abuse material (CSAM) across Facebook and Instagram this year, according to an investigation published Tuesday by the Tech Transparency Project (TTP)2,4. The majority of those ads, 274, ran in August. Many promoted AI-powered "nudify" apps, predominantly from Chinese developers, that digitally alter images of children.
TTP matched multiple ads to photos of real children found online. One ad used an image of a young member of a European royal family, animating it into a video depicting a graphic sex act. Another took a photo of a 14-year-old Instagram influencer in a sports club uniform and morphed it into explicit content. A third depicted a preteen girl from stock photos being molested by an adult male.
The ads ran across the United States, United Kingdom, EU, Australia, and India. About 84 of the 332 ads ran in India, with 78 of those appearing after the Indian government had already ordered Meta to remove all CSAM-promoting ads and content on Instagram. That order followed a BBC Eye investigation that found Instagram was running paid adverts promoting CSAM in India. India's National Commission for Protection of Child Rights and National Human Rights Commission have said they are examining the matter. Just Rights for Children filed a petition in India's Supreme Court citing the BBC's findings and seeking mandatory reporting requirements for social media platforms.
Meta's ad review system relies primarily on automation, checking images, video, text, audio, targeting, and destination links before approving ads for publication. When TTP reported 55 of the 84 India-targeted ads through Meta's reporting system, Meta responded that 43 of them did not violate its Advertising Standards. For 11, Meta said the content had already been removed. Meta only allows users to report ads while they are live. All flagged ads were removed after TTP contacted Meta for comment on its report.
Meta said in a statement that it does not "tolerate nudify apps or any kind of child exploitation, whether real or AI-generated," adding that "these criminals constantly shift tactics to evade detection". The company said many of the flagged ads had already been removed, most had fewer than 200 impressions, and total ad spend was under $5,000. Meta said its defenses include systems that block violating ads and ongoing monitoring.
TTP reported that new ads continued to appear after it made Meta aware of its findings, sharing details of 17 additional ads, including 11 that ran in India on September 1–2. Several of the ads were placed by Meta's own ad agency partners in China, TTP said.
Separately, the San Francisco City Attorney's Office has asked Meta to explain how the ads repeatedly ran on its platforms3. Meta has claimed the ads fall outside the city's jurisdiction. The U.S. Justice Department has clarified that AI-generated CSAM is treated as equivalent to traditional CSAM under federal child pornography law.
ANALYSIS The 43 ads that Meta's own reporting system deemed compliant with its Advertising Standards, despite containing material TTP identified as CSAM, point to a gap between Meta's stated prohibition on child exploitation content and the operational performance of its automated review pipeline. The concentration of 274 ads in a single month, and the continued appearance of new ads after TTP's disclosure, suggest the moderation system's detection capabilities are not keeping pace with the volume of violating content submitted through Meta's paid ad infrastructure.