Meta Approved and Ran Paid Ads Featuring AI-Generated CSAM, Researchers Find

Watchdog group the Tech Transparency Project found more than 50 paid image and video ads containing AI-generated child sexual abuse material in Meta’s own ad library, all of which were reviewed and approved by Meta before running across Facebook, Instagram, Messenger, and Threads between November 2025 and early August. Because these were advertisements rather than user posts, they passed through Meta’s ad-review pipeline — which the company says relies primarily on automated tools — while Meta collected the ad revenue. Several ads linked out to ‘nudify’ or face-swap apps, including MaskAI, which Apple pulled from the App Store after WIRED inquired. Many of the disclosed advertisers traced back to Chinese firms, one of them a former Meta ad reseller.

The reach was mostly small, but at least one ad hit more than 2,500 accounts across Europe, and Meta’s library omits US performance data, so true exposure is likely higher. Notably, some ads went live only after WIRED first questioned Meta, and researchers found duplicates of ads Meta had already removed for policy violations — suggesting the company’s systems can fingerprint known abusive content yet still let identical versions run. This is the second recent case of CSAM-linked ads on Meta’s platforms, following a July BBC investigation into similar ads in India.

Meta removed the ads after being contacted, said most had minimal reach, and pointed to newly deployed AI detection at upload and the 36 million pieces of exploitation content it removed last year. But the incident exposes a structural gap: paid, pre-reviewed content is the one category Meta explicitly vets before publication, and here the vetting failed repeatedly — including against material the company had already flagged. It also highlights how consumer-facing generative ‘undressing’ apps are now being openly marketed through mainstream ad networks.