After researchers alerted Meta to hundreds of ads featuring AI-generated child sexual abuse images that were running on its platform, the ads continued to appear despite the company claiming its new detection tools would block such content.
The issue first came to light when Wired reported last month that about 50 ads running on Facebook, Instagram, and Threads directly contained child sexual abuse material. Meta reportedly removed the ads and said most had been published before it introduced new AI tools intended to detect and block harmful advertising.
Wired reported Tuesday that when researchers at the Tech Transparency Project tried to verify whether the new safeguards were working, they instead found more than 250 additional ads containing child sexual abuse material on Facebook, Instagram, Threads, Messenger, and Meta’s broader advertising network since the beginning of August.
Unlike the initial batch, some of the new ads reportedly used images of real children. The ads typically began with ordinary photographs of preteen or teenage girls before transforming their faces into graphic sexual videos and directing users to AI image- and video-generation apps.
According to Wired, researchers were able to identify four real minors whose images had been used, including three in the US. One was a teenage social media influencer whose photograph had been taken from a public account on one of Meta’s platforms, while another was also a teen influencer. A third was a real person whose photograph appeared on a stock-image site and was labeled as a preteen.
Another ad used an officially released photograph of a member of a European royal family, transforming it into a video depicting a graphic sex act. TTP did not identify the royal to protect the victim, but said it contacted officials in the country.
TTP ultimately found more than 350 abusive video ads dating back to last year, with about 300 promoting AI image- and video-generation apps, many linked to Chinese developers, per Wired. The ads reached tens of thousands of accounts, and researchers reported them to NCMEC and Meta.
Meta removed about 150 ads after TTP shared its findings, and updated more than 100 others to reflect child-exploitation violations. But researchers continued finding dozens of CSAM-containing ads in the days afterward.
Wired reports that many ads directed users to AI editing apps on Apple’s App Store and Google Play, while others appeared tied to fake Facebook accounts and Meta’s Chinese advertising resellers.
David Pham, a former Meta employee, said bad actors continually adapt to new safeguards.
“Every time the platform or investigators would build a new detection model or a new way to have some type of enforcement, they quickly — within a day, maybe five days — built a new way to avoid detection,” Pham told Wired.
Meta spokesperson Tracy Clayton said the company does not tolerate child exploitation and said most of the ads had fewer than 200 impressions, with less than $5,000 in total ad spending. Meta said some were difficult to detect because minors appeared only briefly or were blurred.
The findings add to scrutiny of Meta’s child-safety practices, as the company has faced dozens of lawsuits over alleged harms to children and a New Mexico ruling requiring stronger CSAM reporting procedures. They also come amid growing concerns that AI “nudify” tools are increasingly being used to create CSAM from ordinary photographs of children.
Related: Meta Ordered to Pay $567M, Implement Safeguards as Part of New Mexico Child Mental Health Trial