Meta Failed to Catch Hundreds of AI Child Abuse Ads. Some Included Images of Real Kids
Developed over time first seen 2 months ago
An investigation by the Tech Transparency Project (TTP), shared with WIRED, found that Meta's platforms hosted hundreds of AI-generated child sexual abuse material (CSAM) adverts, despite the company's claims that new AI moderation tools would catch such content. This follows an earlier WIRED report that prompted Meta to delete around 50 similar ads last month, and raises serious concerns about the effectiveness of the company's ad-review systems, which it says vet all content before publication, and about the length of time harmful ads remained live and generating revenue for Meta.
TTP identified more than 250 further CSAM ads posted since the start of August, and over 350 in total since the end of last year, some running for up to a week and racking up hundreds of views after being reported. Unlike earlier ads, some of the new ones used images of real, identifiable children, including a minor from a European royal family whose official photograph was manipulated into an explicit video, and researchers say they traced the real identities of four affected minors, three in the US. The ads, spread across Facebook, Instagram, Messenger and Threads, typically morphed innocuous photos of children into sexual videos before directing viewers to download "nudification" apps, many linked to Chinese developers; Meta reportedly failed to label some removed ads as CSAM violations until WIRED queried the discrepancy.
- Sources
- 2
- Outlets
- 2
- Left · Centre · Right
- 0 · 1 · 0
In short
- Meta hosted 250+ new AI child abuse ads since August, TTP found.
- Some ads used real children's images, including a European royal.
- Reported ads took up to a week to be removed by Meta.
How this story developed
- 2 months ago New reporting shows Meta's response to AI-generated child sexual abuse material (CSAM) on its platforms has been slower and less effective than the company claimed.
- 2 months ago Meta has continued to publish hundreds of AI-generated child sexual abuse ads on Facebook, Instagram and Threads, despite claiming new detection tools would prevent this, according to researchers at the nonprofit Tech Transparency Project (TTP).
- 2 months ago An investigation by the Tech Transparency Project (TTP), shared with WIRED, found that Meta's platforms hosted hundreds of AI-generated child sexual abuse material (CSAM) adverts, despite the company's claims that new AI moderation tools would catch such content.
Explained simply
Meta runs adverts on its platforms, including Facebook and Instagram, and is meant to screen them before they go live. Researchers, including a group called the Tech Transparency Project, have been checking whether that screening actually catches harmful content, and reporting their findings to the outlet WIRED. This latest work focuses on adverts that used artificial intelligence to turn ordinary photos of children into sexual abuse material, then pointed viewers towards "nudification" apps, tools that digitally strip or alter images of real people without their consent.
The concern isn't only that this material appeared, but how slowly it was dealt with and whether Meta's own account of its response holds up. Earlier reporting had already pushed Meta to remove dozens of similar adverts, but researchers say hundreds more surfaced afterwards, some using photos of identifiable real children. Meta's ad-review systems, and newer tools built to detect AI-generated abuse content automatically, are central to the story because they are meant to stop this kind of material before the public ever sees it.
This matters because it touches on child safety, the accountability of a major technology company for what it allows to be advertised, and the wider spread of AI tools that can manipulate images of real people. Apple and Google are also relevant, since the apps promoted in these adverts are distributed through their app stores.