Meta Removed Ads That Promoted AI-Generated Child Abuse Imagery

Meta has removed dozens of paid ads that researchers say displayed AI-generated child sexual abuse imagery and other sexually explicit material involving minors, after they were found running across Facebook, Instagram, Messenger and Threads in the U.S., U.K. and multiple European countries. The findings matter because they show that harmful ads can still pass Meta’s review systems, reach real users, and remain visible in the company’s transparency library for weeks or months.

The ads were identified by the Tech Transparency Project after a new sweep of Meta’s ad database, adding to a growing body of evidence that child-exploitation content can appear in paid placements even as platforms publicly pledge stricter enforcement. Meta says the ads were disabled for policy violations and that it is improving its automated detection tools.

Editor’s note: This article discusses child sexual abuse material and sexual exploitation. Reader discretion is advised.

What researchers found inside Meta’s ad library

Researchers say they uncovered more than 50 image and video ads that violated Meta’s rules by showing or implying child sexual abuse material, nudity, or sexually suggestive depictions of minors. The ads were published between November last year and early August and, in some cases, reached thousands of accounts before being taken down.

According to the Tech Transparency Project, the ads were not hidden in user-generated posts or private messages. They were paid placements that appeared to have been reviewed and approved by Meta’s advertising systems, then catalogued in the company’s public ad library.

That distinction is important. A normal user post can be removed after moderation, but ads are supposed to pass a pre-publication screening process. The watchdog group says the offending creatives often stayed accessible in the ad archive for months even after they were no longer active.

How the ads appeared

One ad, researchers say, used a thumbnail of a child on the floor and overlaid it with text promising “deep fantasies” through AI. When clicked, it reportedly opened a video that began with adult sexual content and then inserted the child’s face into the explicit scene.

Another recurring ad allegedly showed a young girl posed in a sexualized way with copy implying the viewer could be shown more. Other ads featured images of minors alongside language associated with nudify or undressing apps, which use AI to create or simulate sexualized imagery from ordinary photos.

“These ads made no effort to mask the images or hide what they were promoting,” the Tech Transparency Project’s Katie Paul said, arguing that the issue was not third-party organic content but paid advertising that Meta reviewed, approved and sold.

The researchers say several ads were visible only to narrow audiences, including some targeted solely at men. But at least one ad reached more than 2,500 people in Europe, with recipients in France, Germany, Ireland, Italy, the Netherlands, Spain, Sweden and the United Kingdom. Meta’s public ad library does not provide full reach data for every market, so the overall exposure could be higher.

Why this is a bigger problem than one takedown

The latest removals highlight a broader enforcement problem for platforms confronting AI-generated abuse imagery. Even when companies build policies banning nudity, sexual exploitation and child abuse material, the ads ecosystem can move faster than automated moderation.

Researchers and child-safety advocates say the underlying business model makes abuse easier to scale. A single advertiser can create multiple accounts, publish similar creatives, and route traffic to new domains or apps whenever one campaign is blocked. That means enforcement often becomes reactive rather than preventive.

Meta has publicly emphasized its investment in automated review systems. In this case, however, the company’s own ad archive suggests some of the offending placements were approved and allowed to run before being removed. The researchers also say they found apparently identical ads that had previously been taken down, then later resurfaced in new form.

How Meta says it reviews ads

Meta’s advertising standards state that ads are screened before publication, primarily through automated tools, and that child sexual exploitation material is prohibited. The company also says it reports known violations to the National Center for Missing and Exploited Children’s CyberTipline.

After being contacted about the findings, Meta removed the ads from its library and said it works aggressively to keep sexual exploitation off its services. The company said many of the ads had little reach, and some had already been disabled before the report was shared with it. Meta also pointed to a newer AI-based detection system designed to catch violating ads earlier in the upload process.

Meta said it has removed tens of millions of child sexual exploitation items in the past year and has also taken legal action against some nudify app developers as part of its broader enforcement effort.

Still, the company’s response underscores a tension that has defined platform safety work for years: faster AI-generated abuse content means faster AI detection must keep pace. Critics say Meta’s enforcement has historically lagged behind the speed at which abusive advertisers adapt.

Who was behind the ads?

In many cases, the researchers say the ads were linked to accounts with no followers or very small audiences. When an advertiser was identified, some appeared to be connected to Chinese companies or Chinese ad-reselling networks.

One example involved an advertiser called Meet Social, a Chinese firm that the researchers say was tied to an ad showing a sexualized image of a girl reclining. Meet Social did not respond to requests for comment. The company has previously been described in reporting as a major Facebook ad reseller serving Chinese clients despite Meta’s platforms being blocked in China.

The pattern matters because advertiser identity often becomes part of the enforcement problem. Low-signal accounts, foreign resellers and constantly changing web domains can make it difficult to determine whether a campaign is legitimate marketing or an attempt to monetize abusive content.

What happened to the linked apps?

Some of the ads directed users to apps or websites built around nudification tools. One app identified by researchers, MaskAI, appeared in Apple’s App Store but did not display explicit material until after installation. Once opened, it reportedly offered AI-generated pornography and tools for uploading photos and face-swapping people into sexual scenarios.

After being contacted by reporters, Apple removed MaskAI from the App Store. An Apple spokesperson said the company prohibits apps designed to generate, distribute or consume pornography and has strict rules against nudification apps.

The takedown is notable because app-store controls often operate as a second line of defense. If platform ads can push users toward sexual exploitation tools and the app marketplace then hosts them, the abuse pipeline can span multiple companies with different moderation standards.

How did this keep happening after prior warnings?

The most troubling part of the latest disclosure is that it arrived amid repeated alerts about AI nudify advertising on Meta. In recent months, independent researchers, journalists and safety groups have documented a persistent stream of similar ads, suggesting the problem is not isolated.

Earlier this summer, a BBC investigation reported that Instagram ads in India appeared to promote the sale of child sexual abuse material through Telegram channels. Meta responded at the time by restating its zero-tolerance approach to child exploitation. The new findings suggest that formal policy language has not yet translated into consistently effective enforcement.

Researchers say that once a creative format works, it spreads quickly. The same general style of imagery can be re-uploaded with minor edits, new text overlays or new destination links. That means individual takedowns may provide only temporary relief.

Why the ad library matters

Meta’s ad library is designed to increase transparency by showing what ads have run across its apps. In practice, researchers say it can also expose how long harmful ads remain visible after they stop running.

The Tech Transparency Project says some ads stayed in the database for months even after they were no longer active. That makes the library useful for accountability research, but it also reveals a limitation: users can see what was approved, yet there is no obvious way to report a listing once a campaign has ended.

For safety advocates, that gap is significant. If an ad is already gone but remains searchable, the platform still has a record of what passed review, yet the public may not have a direct path to flag it for further action.

Timeline of the Meta ad findings

The following table summarizes the main milestones in the latest reporting and response cycle.

What advocates and researchers are saying

Safety researchers say the case fits a familiar pattern: offensive content is discovered by outsiders, escalated through public reporting and then removed after a delay. By the time platforms respond, the ads may already have been delivered to real users.

Alexios Mantzarlis, a former trust-and-safety worker at Google and co-founder of Indicator, says he has reported tens of thousands of nudifier ads on Meta over several years. He argues that the scale of the ecosystem shows how hard it is for platforms to keep up when malicious advertisers can quickly recreate accounts and websites.

Mantzarlis said image-based sexual abuse has evolved from a fringe abuse tactic into a lasting feature of the online environment, in part because enforcement across platforms has been too cautious and too slow.

Other groups that work on online abuse offered a more mixed picture. The Tech Coalition, which coordinates safety work among companies, said Meta has shared thousands of violating URLs through its Lantern system, which helps companies exchange abuse signals. The U.K. Revenge Porn Helpline also pointed to Meta’s work with StopNCII.org, a tool that helps block nonconsensual intimate images.

Those comments suggest Meta has built partnerships with external safety organizations, but researchers say partnerships alone do not solve the core issue if harmful ads still pass review in the first place.

What are nudify apps and why are they controversial?

Nudify apps are AI tools that claim to remove clothing, simulate nudity or create sexualized versions of real people from ordinary photos. In practice, they are frequently used for harassment, nonconsensual image abuse and, in the most severe cases, the creation of AI-generated child sexual abuse material.

They have become one of the clearest examples of how generative AI can be used for exploitation rather than creativity. The services often market themselves with euphemisms, but their outputs and advertising can be explicit. Once distributed through app stores, social media ads or Telegram channels, they can reach teenagers, schools and non-consenting adults with alarming speed.

• They often use ordinary photos uploaded by victims or users.
• They can generate sexualized images without consent.
• They are repeatedly rebranded after enforcement actions.
• They can intersect with child abuse imagery, raising criminal concerns in many jurisdictions.

What happens next for Meta?

Meta is likely to face renewed pressure to explain how ads featuring abusive or sexualized images of minors were approved in the first place. The company has said its new AI systems are intended to improve detection, but the latest findings suggest the enforcement gap remains serious.

For lawmakers and regulators, the case may become another example of why platform self-policing can fall short when safety risks are embedded in the ad business itself. For parents, child-safety groups and users, it is a reminder that harmful material can appear in places widely assumed to be tightly monitored.

The larger lesson is uncomfortable but clear: AI-generated abuse is no longer just a fringe content problem. It is now part of the mainstream moderation challenge faced by the biggest technology platforms, and the ad systems that monetize those platforms are one of the weakest links.

Meta says it will keep fighting the problem. Researchers say the question is whether the company can do so fast enough to stop the next wave of ads before they are approved, delivered and archived for all to see.