Loading page…
The September 8 investigation says advertisements promoting AI face-swapping and video apps remained on Meta platforms, sometimes using photos of real children.
Listen to this story
The Tech Transparency Project says Meta’s Facebook, Instagram, and Threads carried hundreds of ads promoting AI tools that generated child sexual-abuse material, including content using real children’s photos. Its 2026 count reached 332 Facebook and Instagram ads, with additional cases continuing after August. Meta removed identified ads and reported them to the National Center for Missing and Exploited Children, but some remained live for days and one grew past 330 views.
TTP found more than 350 abusive video ads across Meta platforms since the end of 2025.
The ads reached over 29,000 EU accounts and roughly 7,000 UK accounts, according to Meta ad-library data.
More than 250 additional ads were identified after August began, separate from TTP’s 332-ad 2026 count.
Paid advertisements directing people to AI face-swapping and video apps continued to reach Meta users despite the platform’s detection efforts. In a September 8 investigation, the Tech Transparency Project said Meta failed to detect 332 ads containing AI-generated child sexual abuse material on Facebook and Instagram in 2026; some used photographs of real children.
The 332 figure is TTP’s count for 2026, not the full period researchers examined. The group said it found more than 250 additional ads containing child sexual abuse material across Facebook, Instagram and Threads since the start of August. It separately said more than 350 abusive video ads had run across Meta platforms since the end of 2025.
TTP said Meta failed to detect 332 ads containing AI-generated child sexual abuse material on Facebook and Instagram in 2026.
Researchers identified more than 250 additional Meta ads containing child sexual abuse material since the start of August 2026.
TTP said more than 350 abusive video ads had been published across Meta platforms since the end of 2025.
Researchers described ads that opened on an innocuous image of a child, then morphed the face into explicit sexual video and directed viewers to AI face-swapping or video applications. TTP matched multiple ads to online photos of real children. WIRED verified that one source image came from an official European royal-family website, without identifying the child.
The reported audience was not trivial. Meta ad-library data showed the ads reaching more than 29,000 accounts in European Union countries and about 7,000 in the United Kingdom, according to TTP. The group said the vast majority of the investigated ads promoted AI applications made in China, including tools that sexually alter images of children.
TTP said Meta sometimes took a week to review reports about live ads containing child sexual abuse material. One reported ad remained active for days, growing from four views to more than 330 before removal. After TTP later flagged more than 300 ads, Meta removed roughly half within hours, but some stayed active for days and others were not initially labeled as violations in the company’s ad archive.
Meta said it removed the identified ads, reported the material to the National Center for Missing and Exploited Children, and received under $5,000 in advertising payments from them. The company said many had already been removed and most had fewer than 200 impressions. It attributed some detection difficulty to clips in which depictions of minors were brief or blurred.
The findings came after Meta said it had deployed improved child-exploitation detection systems when an earlier batch of 53 ads was flagged. TTP later found dozens of reported ads still running and hundreds of new ads appearing afterward, a result that puts the practical effectiveness of that upgrade in question.
Editorial analysis
The consequential failure described here is not simply that harmful content existed online. It was paid distribution: ads had to pass through Meta’s systems and, in some cases, stayed live after being reported. Meta attributes some difficulty to brief or blurred depictions and says criminals seek to evade its defenses. That makes removal speed a concrete test of the system, not just its pre-publication screening. Watch whether Meta provides a clearer escalation path and more consistent archive classifications when child-exploitation ads are flagged.
Loading discussion...
Join the conversation
Explain what response standard platforms should meet.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.