San Francisco Presses Meta Over AI Child Abuse Ads After TTP Report

Updated September 10, 2026 12:53 am

San Francisco’s city attorney has formally demanded that Meta stop letting paid ads promoting AI-generated child sexual abuse content run on its platforms, escalating pressure on the company after researchers found hundreds of such placements across Facebook, Instagram, and Threads. The move matters because it turns a disturbing content-moderation failure into a legal and regulatory fight over how a major platform polices abuse, responds to reports, and profits from harmful advertising.

The letter from City Attorney David Chiu comes after research from the Tech Transparency Project found more than 350 Meta ads in recent months that used manipulated images of minors to market AI image and video tools. Some ads appeared to transform still photos of real children — including, in one case, a member of a European royal family — into sexualized video clips, while others steered users toward apps that can create nonconsensual intimate images or “nudify” content.

What did San Francisco demand from Meta?

San Francisco asked Meta to explain how those ads were allowed to run, how the company reviews and escalates potentially illegal content, and what it does when advertisers repeatedly submit abusive material. The city’s demand is not just for a takedown review; it is asking Meta to account for the mechanics of its ad moderation system and the legal obligations that come into play once the company is aware of apparent child sexual exploitation.

In a four-page cease-and-desist letter, Chiu said the office wants Meta to stop “allowing” ads containing AI-generated child sexual abuse content to appear on its platforms and to meet promptly with the city’s legal, child-safety, and advertising teams. The letter also asks for an explanation of how Meta reports such material to the National Center for Missing and Exploited Children, or NCMEC, and how it handles repeat offenders behind the ads.

“No company—particularly one that claims all advertisements are reviewed and approved prior to distribution and then accepts payment for that distribution—should permit its advertising systems to be used in this manner,” the city attorney wrote in the letter.

Chiu also said the evidence suggests the issue may not be an isolated mistake that was fixed once Meta learned about it. Instead, he said the repeated appearance of these ads points to a broader, systemic failure in detection and enforcement.

How were the ads discovered?

They were identified by the Tech Transparency Project, which tracked paid promotions across Meta’s ad ecosystem and documented a pattern of abusive content that slipped through the company’s systems. The ads were not random one-offs. According to the researchers, some of the same ads were uploaded multiple times, and many of them were still circulating even after earlier coverage brought the issue to public attention.

WIRED first reported in early August on a batch of 53 such ads. Since then, researchers found that more than 250 additional ads ran on Meta’s platforms, bringing the total documented count to more than 350. In aggregate, the ads reportedly reached more than 29,000 accounts in European Union countries, and they also targeted users in the United States, Australia, and India.

Some of the ads directed users toward apps that could be used to digitally undress people, create synthetic nudity, or generate sexualized video content from still images. In several cases, the manipulated images involved minors. That is the most serious aspect of the case: the ads were not merely promoting objectionable software, but leveraging content that appears to sexualize children.

Why does this case matter beyond one platform?

This case matters because it shows how AI abuse tools can be monetized through mainstream advertising, turning a moderation problem into a distribution problem. If a platform accepts payment for ads that promote tools used to create abusive imagery, critics argue it is not simply failing to remove harmful content after the fact — it is helping deliver that content to users at scale.

The issue also highlights a gap between platform policy language and real-world enforcement. Meta’s public rules prohibit child sexual exploitation, adult sexual exploitation, nudity, and the promotion of nudify apps. The company says ads are reviewed using automated systems and, in some cases, human review. But the persistence of these ads suggests those controls can be bypassed, whether through repeated uploads, weak detection, or delayed enforcement after reports are filed.

For regulators and city attorneys, the broader concern is that AI-generated abuse content can spread quickly through everyday ad systems that were designed for commerce and engagement, not for identifying synthetic sexual exploitation. That creates new legal and operational questions for platforms that monetize advertising while claiming to police it carefully.

What Meta has said so far

Meta says it has removed the ads found by researchers and that they violated the company’s policies. The company also says there is no evidence the ads were displayed in San Francisco, which it argues places the matter outside the city attorney’s jurisdiction.

Meta has previously said many of the ads uncovered by TTP had already been removed before the researchers reported them, and that most received fewer than 200 impressions. The company also said the combined ad spend for the 300-plus ads was under $5,000. In its earlier response, Meta said it does not tolerate nudify apps or child exploitation, whether real or AI-generated.

Meta has maintained that it does not allow nudify apps or any form of child exploitation on its platforms, and it says its ad review systems are designed to catch violations before delivery. The company says the offending ads were removed and that the apparent reach of many of them was limited.

Those assurances, however, have not satisfied researchers or local officials, especially because new ads reportedly appeared even after the problem was first publicized. TTP says some recently observed ads featured the same children it had already flagged to Meta.

What does the city attorney want to know?

San Francisco wants a detailed explanation of how the ads escaped Meta’s moderation systems, how quickly the company reacts to complaints, and what it does to stop the same advertisers from returning with fresh material. The letter also asks Meta to clarify how it identifies, escalates, and reports material that may involve child sexual abuse to NCMEC, which serves as a central reporting and coordination body in the United States.

In addition, the city wants information on the role of Meta staff in legal, safety, and advertising functions when these ads are detected. The underlying question is whether Meta’s current workflow is too slow, too fragmented, or too dependent on after-the-fact reporting to prevent repeat abuse.

Chiu’s office said it reviewed many of the ads identified by TTP and believes the pattern raises not only operational but also legal concerns. The city is seeking a response from Meta within 28 days.

Key issues raised in the letter

• How paid AI abuse ads were approved and served on Meta’s platforms
• Why some reported ads remained live for more than a week
• How Meta identifies repeat offenders and malicious advertisers
• How the company reports suspected child sexual exploitation to NCMEC
• Whether Meta’s takedowns are reactive rather than preventive

How long were the ads allowed to stay up?

In some instances, the ads remained visible for more than a week after researchers reported them, according to TTP. During that time, the ads were seen by additional users, which magnified the harm and increased the chance that new audiences would be exposed to abusive content.

The delayed response matters because the value of an ad platform depends on speed. Once a harmful campaign is identified, every extra hour it stays online can mean more impressions, more clicks, and more revenue. That is why critics argue that moderation delays are not just technical errors; they can become a business and safety problem at the same time.

TTP also said that when some of the ads were removed, Meta did not always classify them as child sexual abuse content in its ad library. That apparent mismatch between the nature of the content and the platform’s labeling raises questions about how Meta internally categorizes abuse and whether its public records reflect the seriousness of the violations.

What is Meta’s policy on nudify apps and exploitation?

Meta’s policies prohibit child sexual exploitation, child abuse, nudity involving minors, and the promotion of adult exploitation or nudify applications. The company also says ads are reviewed for policy compliance before distribution, using a combination of automated and manual checks.

On paper, those rules appear strict. In practice, researchers have repeatedly found that harmful synthetic-image ads still slip through. Meta has also said it has removed hundreds of thousands of ads tied to nudify apps, suggesting the problem is large enough to require sustained enforcement rather than one-time intervention.

Still, the continued appearance of abuse-related promotions indicates that bad actors can exploit loopholes or overwhelm the system by re-uploading content, changing creative details, or using new advertiser accounts. This is one of the core challenges facing every major platform that sells ad inventory at scale.

Why enforcement is so difficult

AI-generated abuse ads are difficult to police because they may be only slightly altered from one upload to the next, and the content may not be explicit in the first frame or thumbnail. In addition, platforms are forced to balance speed, scale, and false positives, which can lead to missed violations if detection thresholds are too conservative.

Advertisers can also recycle the same creative under multiple names, making takedowns feel like a game of whack-a-mole. That is why critics say enforcement has to focus not only on individual ads, but on the networks, accounts, and payment channels that make repeated abuse possible.

How did the issue spread across Meta’s apps?

The ads ran across Facebook, Instagram, and Threads, according to the reporting. That cross-platform spread matters because it suggests the issue was not confined to one product or one audience segment. Instead, it traveled through Meta’s broader advertising infrastructure, which is designed to distribute content efficiently across multiple surfaces.

That distribution model is effective for legitimate advertisers, but it becomes dangerous when the content itself is harmful. A campaign that reaches tens of thousands of users across different apps can normalize abusive AI tools, drive traffic to exploitative services, and increase the availability of software that can be used to humiliate or blackmail victims.

The international reach is also notable. While the ads reportedly targeted people in the United States, Australia, and India, more than 29,000 accounts in the European Union were exposed. That makes the issue relevant not only to local prosecutors in San Francisco but also to global regulators and child-safety advocates.

Timeline of the Meta ad controversy

The sequence below shows how the story developed from the first report to the city attorney’s legal response.

Why are researchers still finding new ads?

Researchers say the problem persists because Meta’s enforcement is still too slow, too fragmented, or too easy to evade. Katie Paul, who directs the Tech Transparency Project, says the evidence shows the company is not adequately addressing the issue and that new ads have continued to appear even after the latest report was published.

Paul said that after TTP’s September 8 report, brand-new ads still surfaced featuring some of the same children the group had already flagged to Meta. She argued that the continued appearance of such content shows the platform is not effectively solving the problem.

That is especially concerning because it means public reporting alone is not enough to shut down abusive campaigns. If new ads can be launched after a prior takedown, the platform may be chasing incidents rather than eliminating the underlying actor or content pipeline.

The researchers have also reported the ads to both Meta and NCMEC. Some of the ads were removed, but the continued appearance of new material suggests the system is still vulnerable to repeat abuse.

What are the wider implications for AI companies and advertisers?

This case could have implications beyond Meta. It exposes a growing tension in the AI economy: companies that build image and video tools can benefit from aggressive advertising, but those same tools can be used for sexual exploitation, harassment, and deepfake abuse.

That means the ad market may become one of the main battlegrounds for AI safety. If platforms allow deceptive or exploitative promotions to run, they not only enable harmful products but also help normalize them to millions of users. For advertisers, the case is a warning that platforms may face more scrutiny over what they sell and to whom they sell it.

For lawmakers and city attorneys, the case may also serve as a test of how existing child-exploitation laws apply to synthetic media. The law often moves slower than technology, but the harms caused by AI-generated sexual imagery are already affecting real people, especially children and public figures whose likenesses are used without consent.

Potential consequences if Meta does not respond

1. Further legal scrutiny from local or national authorities
2. More pressure on Meta to tighten ad review procedures
3. Greater use of transparency reports to track abusive campaigns
4. Possible policy changes affecting AI image and video advertisers
5. Expanded debate over platform liability for synthetic abuse content

What happens next?

Meta has 28 days to respond to the city attorney’s letter. The company will likely need to explain not only how these ads appeared, but also what it will do differently if similar campaigns are uploaded again. The answer may determine whether the episode remains a public relations problem or becomes a more serious legal confrontation.

For now, the story underscores a hard truth about platform safety: rules against child exploitation are only as effective as the systems used to enforce them. If paid ads can repeatedly slip through, the policies on paper matter far less than the reality on the ground.

As regulators, researchers, and child-safety advocates press for accountability, Meta now faces a larger question than whether a single batch of ads was removed. It must show whether its ad ecosystem can prevent synthetic child sexual abuse material from being bought, served, and spread in the first place.

In short: San Francisco is demanding answers from Meta after researchers documented hundreds of paid ads promoting AI-generated child sexual abuse content across Meta’s platforms, a case that could reshape scrutiny of AI abuse, ad moderation, and platform liability.

Frequently asked questions

What did San Francisco order Meta to do?

San Francisco ordered Meta to stop allowing paid ads that promote AI-generated child sexual abuse content and asked the company to explain how those ads repeatedly ran. The city also wants details on moderation, reporting to NCMEC, and handling repeat offenders.

How many abusive ads were found on Meta’s platforms?

Researchers identified more than 350 ads over recent months. The campaigns appeared on Facebook, Instagram, and Threads, and they reportedly reached more than 29,000 accounts in the European Union, along with users in the US, Australia, and India.

Did Meta remove the ads?

Yes, Meta says the ads were removed for violating its policies. However, researchers and the city attorney argue that removal after the fact is not enough, especially because new ads reportedly continued to appear even after the problem was first reported.

Why is this situation legally significant?

It is legally significant because it raises questions about whether Meta knowingly distributed harmful content through paid advertising and whether it met obligations tied to child exploitation reporting. The city attorney says the pattern may reflect broader systemic failures, not just isolated mistakes.

Are these ads related to “nudify” apps?

Yes, many of the ads promoted AI image- and video-generation apps that can be used to create nonconsensual intimate imagery or digitally undress people. Meta says it does not permit nudify apps, but researchers continue to find ads for them on the company’s platforms.