Meta's New AI Tools Failed to Stop Hundreds of Child Abuse Ads

Researchers at the nonprofit Tech Transparency Project (TTP) say more than 250 new advertisements containing child sexual abuse material have appeared across Meta's platforms — including Facebook, Instagram, Threads and Messenger — since the start of August. The discovery comes shortly after WIRED reported on roughly 50 similar ads that Meta removed, at the time saying it had rolled out new AI tools meant to catch such content.

Real children targeted

Unlike the earlier batch, the new set of ads includes images belonging to identifiable real children, among them a minor from a European royal family whose official photo was turned into an explicit video. Researchers also traced several US cases, including two teenagers with social media accounts and a stock-photo model who had been listed as underage on modeling sites.

In total, Meta has hosted more than 350 such video ads since late last year, many directing users to so-called 'nudification' apps frequently tied to Chinese developers. The ads typically pair an ordinary photo of a young girl with text encouraging downloads of face-swapping apps. Collectively, the ads reached tens of thousands of accounts across the EU, UK, US, Australia and India.

Response from Meta and authorities

A Meta spokesperson said the company does not tolerate exploitation of children, whether real or AI-generated, and that many of the flagged ads had already been taken down, generating under $5,000 in revenue. Apple and Google said most of the linked apps had already been removed from their stores.

The findings have drawn scrutiny from lawmakers and regulators. US Senator Mark Warner urged Meta to halt the ads, while attorneys general in Michigan and Florida, along with Australia's online safety regulator, said they are looking into the matter.