# Meta Ran and Profited From Hundreds of AI Child Abuse Ads for Nine Months > Source: > Published: 2026-09-08 12:38:50+00:00 *Meta's own ad-review system approved and monetised more than 300 ads containing AI-generated child sexual abuse material over nine months, according to a new investigation, and the company only pulled them after researchers and Wired confronted it directly.* These weren't posts that slipped past a moderator on a busy Friday. They were paid ads. Someone at a company bought placement, Meta's automated review approved the creative, and Meta collected money to put it in front of real people. According to a report from the Tech Transparency Project, more than 300 ads containing suspected child sexual abuse material ran across Facebook, Instagram, Threads and Messenger between November 2025 and August 2026, reaching over 29,000 accounts across the US, UK and more than a dozen European countries. Many of the ads promoted deepfake "nudify" apps, tools that strip clothing from photos using AI or generate fabricated sexual images from an uploaded picture. Researchers traced a cluster of them to a Meta advertising partner based in China. Some of the images weren't fabricated composites at all. They were real children's photos, run through AI to generate sexual content. One ad turned an officially released photograph of a minor from a European royal family into a video depicting a graphic sex act. The Tech Transparency Project chose not to name the royal, to protect the child. Meta relies on automated systems to screen ad creative before it ever goes live, and those systems are the same ones that missed this. In some cases, researchers say, it took Meta roughly a week to act on ads they had already flagged as containing CSAM. The ads kept accumulating views the entire time. That's the part that should worry anyone who assumed AI-generated abuse material was rare enough, or crude enough, to get caught automatically. It wasn't. It ran as a normal paid ad campaign, indistinguishable to the system from an ad for shoes. Meta's statement after being contacted was almost identical to statements it has given before: "Sexual exploitation is horrific, and we work aggressively to keep it off our platform." The company also said many of the ads had minimal reach and that some were disabled before Wired shared its findings. That's true in a narrow sense and beside the point in a larger one. The ads existed, they were approved, they ran for months, and Meta's own detection tools weren't what caught them. A nonprofit's researchers and a magazine's reporters were. [Meta Is Testing Robots to Take Over Data Center Grunt Work](https://startupfortune.com/meta-is-testing-robots-to-take-over-data-center-grunt-work/) Meta is piloting robots from Watney Robotics, Kinova and ABB at its Altoona, Iowa and New Albany, Ohio data centers to swap cables, reset servers and cut power. One Meta worker told WIRED the cable-swapping robot could eventually eliminate up to 80% of the workload in some job categories, even as the machines still need close human supervision. - [how to automate data center server maintenance tasks](https://startupfortune.com/meta-is-testing-robots-to-take-over-data-center-grunt-work/) - [robots replacing technician jobs in meta data centers](https://startupfortune.com/meta-is-testing-robots-to-take-over-data-center-grunt-work/) This isn't Meta's first time facing this specific failure this year. In July, Bloomberg reported that Instagram and Facebook had run thousands of ads for nudify apps tied to GatherOne, a Beijing-based firm that connects Chinese advertisers to Meta's platforms. One count put it at roughly 7,600 ads. In early July, the BBC found Instagram serving paid ads in India using search terms like "rape video" that funnelled users to Telegram channels selling illegal material. The pattern repeats. Outside researchers find the ads, Meta removes them, and Meta issues a statement about relentlessly fighting abuse. ## What happens now, and who else is on the hook Apple and Google, whose app stores hosted many of the nudify apps these ads pointed to, responded faster once alerted. Meta didn't. Apple said it has "zero tolerance for developers who deliberately attempt to evade our review process" and removed 17 of 20 flagged apps, with action pending on the rest. Google said it does not allow apps with sexual content and pulled every app named in the report, along with hundreds of others found through its own sweep. Meta was slower. Its response has centred on new detection tooling it says it's now building into ad moderation, a fix promised only after the ads had already run and been monetised for months. Congress is paying attention. Senator Mark Warner has pressed Meta's leadership directly over ads featuring child sexual abuse material and non-consensual intimate imagery, adding political pressure to a story that's already a regulatory one. Lawmakers in the US and Europe have been moving toward mandatory child-safety standards for AI-generated content and platform ad review, and this investigation hands them a case study with a paper trail: a named nonprofit, a named reporting outlet, a nine-month runtime, and a company that says it's committed to fighting exploitation while its own ad system sold space to it. Meta says it removed more than 36 million pieces of child sexual exploitation content last year and has taken legal action against nudify app developers. Both of those things can be true at the same time as this failure. The ad review system is supposed to be the first line of defence, the one gate every piece of paid content has to clear before it reaches a user's feed. For nine months, on this issue, it wasn't a gate at all. **Also read:** [OECD Finds Students Who Use AI Chatbots Daily Score Far Worse on Tests](https://startupfortune.com/oecd-finds-students-who-use-ai-chatbots-daily-score-far-worse-on-tests/) • [A Fake Bookmark Manager Extension Is Turning Chrome Into a Backdoor](https://startupfortune.com/a-fake-bookmark-manager-extension-is-turning-chrome-into-a-backdoor/) • [China's OpenBMB Releases MiniCPM5-2B, Beating Every Open Model Under 4B](https://startupfortune.com/chinas-openbmb-releases-minicpm5-2b-beating-every-open-model-under-4b/) ## Join the discussion [Open in the community →](/community/) Almost there. Sign in and your reply posts straight away.