You post an ordinary photograph of your child online. A sports uniform. A family photo. A picture from a school event. Nothing sexual. Nothing private. Now imagine someone takes that photograph, uses artificial intelligence to transform it into sexual abuse imagery, and then pays one of the world’s largest social media companies to distribute the result as an advertisement.
According to a new investigation, versions of that scenario didn’t just happen once.
On September 8, 2026, the Tech Transparency Project, or TTP, reported identifying 332 paid advertisements in Meta’s advertising system that it classified as containing child sexual abuse material, or CSAM. Most reportedly used AI to transform images of children into sexualized videos.Tech Transparency Project. (2026, September 8). Meta ran hundreds of paid ads with child sexual abuse imagery. https://www.techtransparencyproject.org/articles/meta-ran-hundreds-of-paid-ads-with-child-sexual-abuse-imageryCopy
Investigators traced several of the source photographs to identifiable real children, including young social-media users, a stock-photo model described as a preteen, and a minor belonging to a European royal family.
WIRED, which reviewed TTP’s findings, independently confirmed that one of those source photographs came from an official royal-family website. These children did not pose for sexual images or consent to have their images used or manipulated this way.
Someone allegedly took ordinary photographs of them and turned their likenesses into sexual content, which is illegal and considered child sexual abuse material, CSAM.
Then those images became advertisements.
Researchers Found 332 Ads That Made It Through Meta’s Review Process
Meta says it reviews advertisements before they run, including the images, videos, text, targeting information, and destinations associated with an ad. Yet TTP says it identified 332 ads containing material it classified as CSAM that made it through that process.
The investigation found a recurring pattern. Many ads began with an ordinary photograph of a child and used generative AI or face-swapping technology to transform the image into sexualized content. TTP says it matched several ads to photographs of real children available online. One reportedly came from the Instagram account of a teenage social-media user. Another used stock photographs whose descriptions identified the model as a preteen. Another used the publicly released photograph of the European royal.
Today, a child does not have to create, send, or pose for a sexual image to have their likeness sexually exploited. With generative technology, someone else can manufacture the image for them and apparently use it to run ads through Meta.
Most of the Ads Were Selling Access to More AI Sexualization
The advertisements weren’t simply distributing abusive imagery. Many reportedly functioned as marketing for tools capable of producing more sexualized content. TTP found that 298 of the 332 ads directed users toward AI image- or video-generation applications in Apple’s App Store or Google Play Store. Of those, 253 pointed to apps from China-based developers.
Researchers say 182 ads linked specifically to Apple App Store applications they tested and verified could digitally “nudify” photographs or place uploaded images into sexual videos. That means the alleged exploitation inside the advertisement also served as a demonstration of the product being sold.
A person could encounter an ad containing sexualized imagery of a child, click the advertisement, reach a mainstream app store, download the advertised product, and potentially use it to manipulate another person or a child’s photograph. This is one reason conversations about AI-generated sexual exploitation cannot focus only on the person who ultimately types a prompt or uploads a photograph.
There is an infrastructure around that decision, and software allowing the abuse to occur.
Exploitation Can Move Through an Entire Commercial Pipeline
Think about how many systems can become involved before a synthetic sexual image reaches someone’s screen.A developer builds an application capable of manipulating photographs. To create that application, especially if it’s AI, it had to be trained, most often by similar images. An advertiser then creates promotional content. An advertising network reviews and distributes the ad, and a social platform places it in front of users. An app store hosts the product and makes downloading it easy. Payment systems may process subscriptions or purchases. And the person whose photograph becomes the raw material may never have agreed to any part of it.
Not to mention how many people or coroporations are profiting off of the abuse of these children?
The commercial ecosystem surrounding “nudify” tools has highlighted the same broader problem: nonconsensual sexual imagery can depend on an interconnected network of search engines, social platforms, app stores, developers, payment providers, hosting companies, and marketers.
The new Meta investigation provides an especially stark example because the alleged abusive imagery itself appeared in paid advertising. These weren’t merely posts that users uploaded faster than moderators could remove them. According to TTP, they were advertisements that entered Meta’s commercial advertising infrastructure, passed through review, and were distributed to users.
Meta Had Already Announced New Technology to Fight “Nudify” Ads
The findings are particularly significant because Meta had publicly described its efforts to stop this exact category of advertising.In June 2025, Meta announced that it had developed technology specifically designed to detect advertisements for “nudify” apps. The company said its rules prohibit nonconsensual intimate imagery, including AI-generated material, as well as promotion of apps that create it.
Meta acknowledged that advertisers try to evade detection, including by using seemingly harmless imagery, changing domains, and otherwise disguising the services they promote. The company said it was developing matching technology to identify copycat advertisements and expanding the safety-related language its systems could detect.
Then, in early August 2026, TTP shared an initial group of 53 alleged CSAM advertisements with WIRED.
Meta told WIRED that many of those ads predated new AI technology the company had recently launched to better identify violating advertisements during upload. But TTP says it subsequently identified hundreds of additional abusive ads.
That suggests the problem cannot be reduced to one outdated detection system.
TTP Says Meta Initially Found No Violation in 73 Reported Ads
The investigation also raises questions about what happened after researchers alerted Meta. TTP says it reported 129 active ads through Meta’s reporting system, identifying the reports as involving someone under 18 and apparent sexual exploitation.
According to TTP, Meta initially responded that 73 of those ads did not violate its advertising standards. In other cases, researchers say reviews took days. WIRED reported that one ad involving the photograph of a real child had reached only four accounts in the European Union when researchers reported it. By the time Meta removed it, more than 330 additional users had reportedly seen it.
Content moderation isn’t only about whether harmful material eventually disappears. Speed determines how many people encounter it, save it, click it, or share it somewhere else.
As of September 1, Meta had removed 183 of the 332 advertisements from its Ad Library, according to TTP. After researchers shared the full findings with Meta on September 2, the company removed the remaining 149 within hours. TTP also found that 113 removed ads initially carried violation notices that did not identify child exploitation. Instead, Meta reportedly labeled them under adult-content policies or used generic notices. TTP says Meta changed those notices after researchers raised the discrepancy.
Meta told WIRED that it does not tolerate nudification applications or child exploitation, whether real or AI-generated. The company also said most of the identified ads received fewer than 200 impressions and that advertisers spent less than $5,000 on them in total.
The available reach data still shows that the advertisements reached thousands of people. TTP calculated more than 29,000 accounts reached in the European Union and approximately 6,800 in the United Kingdom. Those numbers do not represent worldwide reach. Meta’s public advertising data does not provide comparable historical reach information in every market, meaning researchers could not determine the total number of people who encountered the ads globally.
Some Ads Were Connected to Meta Advertising Partners
TTP also identified another layer in the commercial system. The organization reported that three of Meta’s advertising reseller partners in China—GIMC, Meetsocial, and BlueFocus—placed advertisements containing the same alleged CSAM content. These reseller companies help Chinese advertisers purchase ads on Meta platforms.
TTP’s investigation therefore raises a broader question about responsibility in digital advertising: How should platforms monitor not only individual advertisers but also the commercial partners that feed advertising into their systems? The issue isn’t simply who created an abusive image, although thats extremely problematic in the first place.
It is also about who helped distribute it, who approved it, who provided access to the software it promoted, and which systems financially benefited along the way.
Apple and Google Also Became Part of the Chain
The advertisements often sent users beyond Meta. TTP says 298 ads linked to AI image or video applications available through Apple or Google. After researchers shared the identified apps with the companies, Apple removed the remaining apps TTP had flagged in its store, while Google told WIRED that it suspended all Google Play apps identified by the investigation.
Apple said some developers initially submitted compliant applications and later introduced prohibited nudification features—a practice sometimes called a “bait-and-switch.” Google said its policies prohibit apps containing sexual content or allowing users to nudify photographs.
Focusing on one platform at a time can become a game of digital whack-a-mole. An advertiser disappears. Another account appears. An app gets removed. A similar one launches under another name. A platform blocks a keyword. Marketers change the language. A domain gets blocked. Operators register another.
The underlying market survives unless companies address the network supporting it.
AI Child Sexual Abuse Imagery Is Growing Rapidly
The Meta investigation comes as child-protection organizations document rapid growth in realistic AI-generated abuse imagery. The Internet Watch Foundation reported that its analysts assessed 8,029 AI-generated images and videos as realistic child sexual abuse imagery during 2025. That included 3,443 AI-generated videos, compared with only 13 such videos recorded by the organization in 2024.
The IWF cautions that AI-generated material still accounts for a relatively small share of the enormous volume of CSAM its analysts encounter. But it says the volume and sophistication of AI material have increased dramatically as the tools become easier to use. Accessibility changes the scale of the problem.
Someone no longer needs sophisticated image-editing skills to manipulate a photograph. Increasingly powerful tools can automate much of the process. And the source image can be completely ordinary.
Children Already Experience Sexual Exploitation Through Mainstream Digital Platforms
The problem also exists within a much broader landscape of technology-facilitated child sexual exploitation.
UNICEF’s September 2026 report Through Children’s Eyes: How Digital Technologies Enable Child Sexual Abuse draws on nationally representative surveys of approximately 21,000 internet-using children ages 12 to 17 across 21 countries, alongside interviews with 100 young people who experienced abuse in childhood. UNICEF estimates that about 20 million internet-using children ages 12 to 17 across those countries—nearly one in five—experienced at least one form of technology-facilitated sexual exploitation or abuse in a single year.
More than half of the reported abuse occurred on mainstream social-media platforms. These findings do not mean every platform causes abuse, nor do they suggest every digital interaction is dangerous. They demonstrate something more useful: child sexual exploitation does not exist only on hidden websites or obscure corners of the internet.
It can move through the same digital infrastructure children and families use every day.
A Synthetic Image Can Still Exploit a Real Child
One of the most dangerous misconceptions surrounding AI-generated sexual content is that a fabricated image has no victim because the depicted event never physically happened. But when an identifiable child’s photograph becomes the basis for sexual imagery, the technology has manipulated a real person’s likeness without their consent.
The event may be fabricated. The child’s identity isn’t. Their loss of control over their image isn’t. And the possibility that strangers can view, save, circulate, or repeatedly manipulate that likeness isn’t imaginary.
That is why discussions about AI sexual content have to remain grounded in human dignity and consent. No photo, publicly accessible or not, gives strangers permission to sexually manipulate the person depicted.
Protecting Children Requires More Than Removing Images Afterward
Takedowns matter. When abusive material appears online, companies should remove it quickly. But by the time an abusive advertisement reaches someone’s feed, multiple safeguards may already have failed. Developers can build stronger protections against sexualizing minors or creating nonconsensual intimate imagery.
App stores can test applications for hidden and post-review functionality rather than relying only on how an app behaves when initially submitted. Advertising platforms can scrutinize creative content, landing pages, advertiser histories, and networks of repeat offenders. Social platforms can enforce policies against accounts and marketers promoting these tools. Payment and commercial relationships can create additional opportunities to disrupt businesses built around abuse. And technology companies can share signals about known abusive services rather than treating each platform as an isolated environment.
Meta itself acknowledged this need in 2025 when it announced that it would share URLs associated with violating nudify services with other technology companies through the Tech Coalition’s Lantern program.
The newest investigation shows why that kind of coordinated approach matters. AI-generated sexual exploitation is not merely a content problem.It is a systems problem. It’s educating on the harms of digitally created exploitation.
What Should You Do If You Encounter Child Sexual Exploitation Online?
People who encounter suspected child sexual exploitation should avoid downloading, saving, forwarding, or redistributing the material. In the United States, suspected online child sexual exploitation can be reported to the National Center for Missing & Exploited Children’s CyberTipline. If someone is in immediate danger, contact local emergency services.
NCMEC also operates Take It Down, a service designed to help prevent the online sharing of nude, partially nude, or sexually explicit images and videos taken of someone before they turned 18.
The Meta investigation should challenge the way we think about responsibility for AI sexual exploitation. Yes, individuals who create and distribute abusive imagery bear responsibility for their choices. But when an entire commercial pipeline makes that exploitation easier to discover, advertise, purchase, generate, and distribute, accountability cannot end with the person holding the phone. Technology companies have built extraordinarily sophisticated systems for identifying what users might click, predicting what they might buy, and delivering advertisements at enormous scale.
Protecting children deserves sophistication at least equal to the systems designed to capture our attention. Whether an abusive image began with a camera or an AI model, whether it appeared on an obscure website or in a paid social media advertisement, the principle remains the same: we are people, not content, not objects for exploitation. Our bodies, identities, and images, especially children’s, should never become products for someone else’s sexual gratification or profit.
Your Support Matters Now More Than Ever
Most kids today are exposed to porn by the age of 12. By the time they’re teenagers, 75% of boys and 70% of girls have already viewed itRobb, M.B., & Mann, S. (2023). Teens and pornography. San Francisco, CA: Common Sense.Copy —often before they’ve had a single healthy conversation about it.
Even more concerning: over half of boys and nearly 40% of girls believe porn is a realistic depiction of sexMartellozzo, E., Monaghan, A., Adler, J. R., Davidson, J., Leyva, R., & Horvath, M. A. H. (2016). “I wasn’t sure it was normal to watch it”: A quantitative and qualitative examination of the impact of online pornography on the values, attitudes, beliefs and behaviours of children and young people. Middlesex University, NSPCC, & Office of the Children’s Commissioner.Copy . And among teens who have seen porn, more than 79% of teens use it to learn how to have sexRobb, M.B., & Mann, S. (2023). Teens and pornography. San Francisco, CA: Common Sense.Copy . That means millions of young people are getting sex ed from violent, degrading content, which becomes their baseline understanding of intimacy. Out of the most popular porn, 33%-88% of videos contain physical aggression and nonconsensual violence-related themesFritz, N., Malic, V., Paul, B., & Zhou, Y. (2020). A descriptive analysis of the types, targets, and relative frequency of aggression in mainstream pornography. Archives of Sexual Behavior, 49(8), 3041-3053. doi:10.1007/s10508-020-01773-0Copy Bridges et al., 2010, “Aggression and Sexual Behavior in Best-Selling Pornography Videos: A Content Analysis,” Violence Against Women.Copy .
From increasing rates of loneliness, depression, and self-doubt, to distorted views of sex, reduced relationship satisfaction, and riskier sexual behavior among teens, porn is impacting individuals, relationships, and society worldwideFight the New Drug. (2024, May). Get the Facts (Series of web articles). Fight the New Drug.Copy .
This is why Fight the New Drug exists—but we can’t do it without you.
Your donation directly fuels the creation of new educational resources, including our awareness-raising videos, podcasts, research-driven articles, engaging school presentations, and digital tools that reach youth where they are: online and in school. It equips individuals, parents, educators, and youth with trustworthy resources to start the conversation.
Will you join us? We’re grateful for whatever you can give—but a recurring donation makes the biggest difference. Every dollar directly supports our vital work, and every individual we reach decreases sexual exploitation. Let’s fight for real love:
