San Francisco City Attorney David Chiu has given Meta 28 days to explain how more than 350 advertisements that turned still images of minors into short sexualized video clips ran on Facebook, Instagram and Threads. His four-page letter to Meta's lawyers followed a WIRED report published Tuesday. Meta's response to the city was jurisdictional: it says it has no indication the ads were shown in San Francisco, which in its view puts the matter outside a city attorney's authority.
The ads were found by researchers at the Tech Transparency Project. Clicking one led to a download for image and video generation apps, some of which digitally undress people or produce other intimate images without consent. Among the minors depicted were real people, including a member of a European royal family. Some identical creatives were uploaded more than once. Together the ads reached more than 29,000 accounts across European Union countries, and the audience also included users in the United States, Australia and India.
This is the second pass at the same problem. In early August, WIRED reported 53 such ads. More than 250 others from the same total kept running after that story published. TTP published its report on 8 September; by Wednesday, its director Katie Paul said, new ads were still appearing, some featuring the same children whose images the organization had already sent to Meta. TTP reported the new ads to Meta and to the National Center for Missing and Exploited Children; some had been taken down by then. Meta did not immediately answer WIRED's question about them.
Chiu told WIRED that Meta knew about AI-generated sexual material in its ad system, did not fix it, and collected revenue while it ran. He wrote that the harm falls on real children, and that the scale and duration of the problem do not match the zero-tolerance policy on child exploitation the company states publicly. His letter leans on Meta's own description of its pipeline: the company says every ad is reviewed and approved before distribution, and is then paid for placement. On that basis, in his reading, the ad system should never have permitted this use. He added that the findings do not look like a single failure the company corrected after the first report.
The letter demands that Meta stop running such ads on its platforms; explain how they got past moderation; describe how it passes ad information to NCMEC; describe how it handles the advertisers who placed the material; and explain what it does about repeat offenders. The office also asked for prompt working sessions with Meta's legal, child safety and advertising staff, and set a 28-day deadline for a reply. Its position is that the findings indicate broader systemic gaps in detecting child sexual abuse signals in advertising, and that they raise serious legal questions for the company.
TTP's researchers found that in some cases Meta did not respond to reports of harmful ads for more than a week, during which hundreds more people saw them. TTP and a check of Meta's ad library also found that some ads, once removed, were not initially labeled as containing child sexual abuse material.
Meta's account is different in emphasis. It told WIRED that many of the ads researchers found had already been removed before journalists asked about them, and that every ad in the sample is now gone as a policy violation. It says most of the ads drew fewer than 200 impressions from accounts, and that the combined advertising budget of the more than 300 ads researchers surfaced came to under $5,000. Its community standards and advertising rules prohibit sexual exploitation, violence and nudity involving children, and prohibit ads for digital undressing apps; all ads, it says, are checked against those rules by automated systems and sometimes by people. It has previously removed hundreds of thousands of ads promoting such apps, sued a Hong Kong company connected to one group of these platforms, and worked with other tech firms on the problem. While reporting on TTP's findings, WIRED found hundreds of new undressing ads still live.
The $5,000 is the number I would sit with, because Meta offered it as reassurance and it does the opposite. Nobody asked what these ads earned. Read as a cost rather than as revenue, it says that placing more than 300 ads in front of more than 29,000 accounts in the EU alone cost less than five thousand dollars — an ad system that is cheap to abuse is a worse finding than one that is lucrative to abuse, because cheap scales without a motive. The fewer-than-200-impressions defense has the same shape: it measures distribution when the letter is about approval. An ad that reaches 199 accounts still passed a review Meta says happens before anything runs.
The part of this nobody is pressing on is the reporting duty. TTP found that some removed ads were not initially classified as child sexual abuse material, and classification is what determines whether a referral goes to NCMEC at all. Chiu asked directly how Meta passes ad information to the center. Everything Meta has said in public concerns takedowns, impressions and budgets. There is no figure for how many of these 350-plus ads were reported, or how quickly. Removal is a content decision; referral is a legal one, and only one of the two has been answered.
Alex Barrett-Shorter of the city attorney's office told WIRED that San Francisco residents can reach the illegal ads through Meta's platforms, which makes it the city's business. He called Meta's reply striking — a company more concerned with whether San Francisco is allowed to raise the subject than with whether children were exploited through its advertising systems.
He has the better of that exchange, and the reason is structural. Meta's ad library discloses some information about ads published in the United States but does not show which parts of the country saw them. The company's defense is that nobody can place these ads in San Francisco, and the record that would settle it is one Meta built and chose not to make granular.