A severe breakdown in advertising moderation has been revealed at Mark Zuckerberg’s Meta, where dozens of paid advertisements containing explicit AI-generated child sexual abuse material (CSAM) and sexually suggestive depictions of minors were approved and displayed over a nine-month period. Research into Meta’s transparency database shows that these sponsored promotions targeted users across the United States, the United Kingdom, and more than a dozen European nations. Crucially, the technology giant collected advertising revenue for serving these illicit materials directly to user feeds.
Uncovering Paid Exploitation Ads Across Meta Services
Researchers at the Tech Transparency Project (TTP), an independent watchdog organization, identified over 50 image and video advertisements containing abusive imagery within Meta’s public ad library. Many of these listings provided direct links to illegal 'nudify' or automated undressing applications. Meta’s ad library serves as an official transparency catalog tracking promotional content across Facebook, Instagram, Messenger, and Threads. This revelation marks the second time in recent weeks that commercial ads featuring child sexual exploitation content have been uncovered operating on Meta’s infrastructure.
Addressing the findings, Katie Paul, director of TTP, emphasized that the advertisements made no attempt to conceal their graphic nature or intent. She highlighted that these were not organic user posts uploaded to social feeds, but paid advertisements that passed Meta's internal review mechanisms, gained approval, and ran without intervention while the corporation generated profit.
Specific Mechanics and Disturbing Examples of Generative Content
The discovered advertisements were published between November of last year and early August. According to Meta’s database records, many ran continuously over multiple days, with several specifically targeting male audiences. Although Meta subsequently removed the listings for violating policies on child sexual exploitation, abuse, and adult sexual solicitation, the advertisements remained archived and accessible in the system for months.
In one documented video advertisement, a thumbnail featured a young child sitting on the floor accompanied by overlay text claiming that users could realize deep fantasies with generative AI. Watchdog analysis showed that clicking the ad initiated a video stream showing adults engaged in explicit acts before using AI face-swapping features to overlay the child's face onto adult bodies.
Another recurring image advertisement featured a young girl reclining in a sexualized pose under the caption claiming she could show more. A third instance involved a video starting with a thumbnail of a girl applying lip gloss, which quickly morphed through generative manipulation into explicit abuse imagery once played.
Lapses in Automated Moderation and Enforcing System Integrity
Meta’s published standards state that all commercial advertisements must undergo review before publication, relying primarily on automated software tools to screen content against corporate policies. These policies explicitly forbid child sexual exploitation and abuse material, requiring mandatory escalation to the National Center for Missing and Exploited Children (NCMEC) when detected. Furthermore, standard ad guidelines prohibit non-consensual nudity, explicit sexual acts, and suggestive poses.
Despite these formal prohibitions, many offending ads remained undetected within Meta's infrastructure. While researchers initially cataloged approximately two dozen violating entries, an additional 30 ads were discovered shortly thereafter. Multiple campaigns remained active and actively served to user accounts even after inquiries were directed to the company.
Furthermore, TTP noted that the dataset contained identical advertisements that Meta had previously removed for policy violations. Katie Paul pointed out that the re-appearance of duplicate CSAM-related ads highlights systemic flaws in Meta’s automated safety filters. She also noted that the public ad library lacks any functional user interface mechanism to flag or report illegal content once an ad campaign finishes running.
Geographic Reach, Target Audiences, and Database Visibility Limits
Although most of the flagged ads were no longer actively serving impressions when cataloged by researchers, they remained stored in Meta's database. Performance metrics indicated that at least one advertisement reached 2,563 individual accounts across Europe, including users in France, Germany, Ireland, Italy, the Netherlands, Spain, Sweden, and the United Kingdom.
Because Meta’s ad transparency database limits granular reach data for the United States and several other international regions, experts warn that the overall exposure metrics could be substantially higher. Following the discovery, TTP formally submitted its findings to NCMEC’s CyberTipline to facilitate law enforcement reporting.
Behind the Commercial Actors: Resellers, Chinese Developers, and App Stores
Investigative data indicates that the primary accounts purchasing these ads possessed minimal or zero social media followers. Where advertiser disclosures were listed, they frequently connected to corporate entities based in China. In the case of the reclining girl advertisement, Meta’s record identified the purchaser as Meet Social, a Chinese marketing firm. Meet Social previously acted as an official ad reseller for Meta, managing ad inventory originating from China, where Meta's consumer platforms are officially blocked. Historical corporate profiles indicate the reseller published thousands of ads daily on Facebook, managing over $1 billion in annual ad sales volume.
Additionally, several promotions funneled user traffic to an application called MaskAI on Apple’s App Store. Developed by a Chinese software creator, the app presented a clean storefront without explicit imagery. However, upon installation, the software provided generative AI pornography tools and face-swapping capabilities designed for sexual contexts. Apple subsequently removed MaskAI from its App Store for violating rules against nudification software, reiterating its strict prohibition on apps designed to generate or distribute explicit content.
Industry Standards, Partner Tools, and Platform Enforcement Debates
Following public scrutiny, Meta cleared the identified listings from its ad library. A company spokesperson reiterated that sexual exploitation is horrific and stated that Meta works aggressively to eliminate it. The company stated that most of the flagged ads had minimal reach, many were disabled prior to external inquiry, and several predated newly deployed AI detection tools designed to intercept violating uploads automatically. Meta highlighted that it removed over 36 million pieces of child sexual exploitation content last year while taking legal steps against nudify app creators.
In connection with industry safety efforts, Sean Litton, outgoing CEO of the Tech Coalition, noted that Meta shares violating URLs through its Lantern data-sharing initiative to help peer platforms identify cross-service abuse. Similarly, representatives from the UK Revenge Porn Helpline acknowledged Meta’s integration with StopNCII.org while stressing that the rapid evolution of nudification services presents an adversarial challenge across digital ecosystems.
However, safety experts argue that enforcement remains insufficient. Alexios Mantzarlis, co-founder of digital deception outlet Indicator and a former trust and safety specialist at Google, revealed he has personally reported more than 25,000 AI nudifier ads on Meta platforms. While Meta reports taking down over 344,000 nudifier ads and launching litigation against Hong Kong entities, Mantzarlis warned that bad actors easily bypass controls by generating fresh ad accounts and domain names, turning image-based abuse into a permanent fixture of online platforms.
