Meta Missed Hundreds of AI Child Abuse Ads, Including Real Kids

Systemic Failures in AI Moderation

Meta’s automated systems failed to block hundreds of advertisements featuring child sexual abuse material (CSAM) across its family of apps, including Facebook, Instagram, and Threads. This revelation follows an initial cleanup early last month when the company removed approximately 50 ads after WIRED reported their existence. At that time, Meta attributed the oversight to a lag between the publication of the ads and the rollout of new AI detection tools intended to filter harmful content. However, subsequent investigations indicate these safeguards were insufficient to prevent widespread exposure.

Researchers from the nonprofit Tech Transparency Project (TTP) have identified more than 250 additional abusive ads running since the beginning of August. Unlike the initial batch, which consisted largely of synthetic media, these newer instances contain verified images of real children. The scope of the violation is significant, with TTP researchers confirming over 350 such video advertisements published on Meta’s platforms since late last year.

Real Victims and Royal Imagery

The content discovered by researchers extends beyond generic AI-generated imagery to include exploitative manipulations of actual individuals. Among the disturbing findings was a video created from an official photograph belonging to a member of a European royal family. The ad transformed this static image into a graphic depiction of a sex act. To protect the minor involved, researchers declined to name the specific individual but confirmed they notified relevant authorities. WIRED independently verified that the source image originated from the official website of that royal house.

Katie Paul, director of the TTP, emphasized the severity of the situation, noting that the platform is profiting from the distribution of this material.

> “What we’re talking about here is not simply AI-generated child sex abuse material,” says Katie Paul, the director of the TTP. The researcher says it appears some images have been taken from the web before being used in the ads. “There’s a lot of harm that can come from this all while Meta is raking in money in the process.”

Link to Exploitative Apps

A recurring pattern in these violations involves links to third-party applications. Many of the abusive ads directed users toward “nudification” apps, primarily developed by Chinese companies. These applications allow users to upload photos of clothed individuals and generate explicit nude versions using artificial intelligence. When clicked, the ads prompted downloads from both Apple and Google app stores. The promotional text within the videos often featured innocent images of teenage girls, overlaid with claims that there were no restrictions on what could be done with the photo. Once the short video played, the faces of the minors were morphed into sexually explicit animations.

Despite Meta’s stated policy that all ads undergo review before appearing on its networks, these violations slipped through for extended periods. In several cases documented by WIRED, it took up to a week for Meta to remove ads that had already been reported via the platform’s internal tools. During this delay, the ads accumulated hundreds of views, exposing unsuspecting users to graphic content.

Lack of Transparency and Accountability

Further complicating the issue is Meta’s handling of the removal process. When the initial wave of CSAM ads was taken down, the platform did not immediately disclose that the removals were due to violations of child safety policies. These details only appeared after direct inquiry from journalists. Additionally, while TTP researchers reported the live ads to US child safety authorities and Meta simultaneously, the delayed response meant the content remained accessible to the public long after it should have been blocked. This incident highlights critical gaps in how major tech platforms utilize AI to protect vulnerable populations from digital exploitation.