Meta Approved Ads With AI-Generated Child Abuse Imagery

Meta's advertising review systems approved and ran more than 50 paid advertisements containing AI-generated child sexual abuse material across Facebook, Instagram, Messenger, and Threads between November 2023 and August 2024, according to findings from the Tech Transparency Project shared with WIRED.

The ads, which researchers discovered in Meta's public ad library, reached several thousand accounts in the United States, United Kingdom, and more than a dozen European countries. At least one advertisement reached 2,563 accounts across multiple European nations. Many of the ads linked to so-called nudify or undressing applications that use artificial intelligence to create non-consensual sexual imagery.

Why it matters

This incident exposes critical gaps in Meta's content moderation systems at a time when the company is investing heavily in AI safety tools. Unlike user-generated content that might slip through automated filters, these were paid advertisements that Meta's review process explicitly approved and profited from. The failure is particularly concerning given Meta's stated "zero tolerance" policy on child exploitation and raises questions about whether advertising revenue creates blind spots in platform safety systems.

How the ads evaded detection

Katie Paul, director of the Tech Transparency Project, told WIRED that the advertisements "made no effort to mask the images or hide what they were promoting." The content was reviewed, approved, and allowed to run by Meta's systems, which the company says rely primarily on automated tools to check ads against its policies before publication.

Some of the violating ads remained in Meta's ad library for months after they stopped running, apparently undetected by the company's monitoring systems. Multiple ads discovered by researchers were published after WIRED initially contacted Meta about the issue, and several were actively showing to users at the time they were found.

Meta acknowledged that some offending ads bypassed new AI detection technology the company recently launched to better identify and block violating content at upload. A company spokesperson said the majority of ads had minimal reach and many were disabled before WIRED shared them, adding that Meta removed over 36 million pieces of child sexual exploitation content last year.

Pattern of identical violations

Researchers identified what appeared to be identical ads that Meta had previously removed for policy violations, only to have them reappear later. Paul said this pattern "raises significant questions about how seriously Meta is taking child exploitation in paid advertisements if its own systems can identify identical ads but it does not remove those identical ads."

The ads were primarily posted by accounts with zero or very few followers. Where advertisers were disclosed, they were linked to Chinese companies, including one firm that was previously a Meta ad reseller handling advertisements in China, where Meta's social platforms are blocked.

Several ads linked to an app called MaskAI that appeared in Apple's App Store. After being contacted by WIRED, Apple removed the application for violating policies against nudification apps. The app exclusively contained AI-generated pornography and included features allowing users to upload images and face-swap people into sexual situations.

Meta removed all the ads identified by researchers after WIRED reached out to the company. The ads now display messages in Meta's ad library indicating they violated policies on child sexual exploitation, abuse and nudity, and adult nudity and sexual activity.

The Tech Transparency Project reported its findings to the National Center for Missing and Exploited Children's CyberTipline immediately upon discovery. This marks the second time in recent weeks that paid ads linked to child sexual abuse material have been found on Meta's platforms, following a BBC investigation in July that uncovered similar content targeting users in India.

These details were first reported by WIRED.