San Francisco Orders Meta to Stop AI Child Abuse Ads

The San Francisco City Attorney’s Office sent a cease-and-desist letter to Meta by September 9, 2026. It demanded that the company stop running paid advertisements containing AI-generated depictions of child sexual abuse. The office also demanded an explanation of how the ads were repeatedly distributed. Reporting by Wired’s Matt Burgess said the series of ads exceeded 350 in number.

The ads in question ran on Facebook, Instagram, and Threads. They included content that turned still images of minors into short videos. Some were reportedly altered into depictions suggestive of sexual acts. Cases involving real people as victims have been confirmed. The alleged inclusion of a member of a European royal family also drew attention. Clicking the ads led users to sources for image-generation and video-generation apps. Some of the apps could reportedly be misused to remove clothing from people. Others could reportedly be used to generate non-consensual intimate images. The situation severely calls into question the effectiveness of ad review and child protection.

More Than 350 Harmful Ads and How They Were

Published

The researchers who discovered the problem were from the Tech Transparency Project. The group identified the ads by examining Meta’s ad library. In some cases, ads with identical content were posted multiple times. In total, more than 350 ads were distributed over several months. Within the European Union, they reached more than 29,000 accounts. Accounts in the United States, Australia, and India were also targeted. Wired first reported a set of 53 ads in early August. Distribution did not stop afterward, with more than 250 additional ads placed. The failure to correct the issue after the initial flagging is viewed as serious. The ads were reportedly created by manipulating images of real children. This is an act that directly harms the dignity and safety of victims. There are suggestions that the ad placement system may have been abused. Under the guise of promoting generative AI apps, they showed harmful edited examples. The tactic was to stimulate users’ curiosity and lead them to install the apps. The volume that evaded review and the length of time involved are abnormal. The possibility of an organized ad operation cannot be ruled out.

Cease-and-Desist Letter Issued by the

City Attorney

The four-page cease-and-desist letter was sent to Meta’s attorneys. The sender was San Francisco City Attorney David Chiu. He made clear that he would not tolerate the continued placement of the ads. He criticized the company for leaving a known problem unaddressed while earning ad revenue. He took seriously the harm caused to real children. He pointed to the gap between its zero-tolerance pledge and the reality. He indicated that current measures are grossly inadequate.

“Meta has failed to address a known issue of AI-generated child sexually explicit ads, while profiting from them,”

He also addressed the premise of the advertising system. Meta has explained that all ads are reviewed and approved in advance. He argued that its responsibility is heavy since it accepts payment for distribution.

“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 letter questioned the operation of ad guidelines and safety mechanisms. It requested prompt discussions with legal, child safety, and advertising staff. It characterized the issue as a structural defect, not a one-off failure. It emphasized the failure to correct the problem after the initial warning. This is now a stage where explanations of recurrence prevention and victim redress are required.

Failures in the Review System and the

Jurisdictional Dispute

Meta pushed back on jurisdictional grounds. It claimed there was no trace of the ads having been shown within the city. It therefore took the position that the matter was outside the City Attorney’s jurisdiction. The ad library includes some U.S. distribution records by design. It does not break down in which U.S. region they were shown. Its structure makes it difficult for outsiders to verify regional display. City spokesperson Alex Barrett-Shorter rejected the rebuttal. He said consumers in the city could encounter the ads in Meta’s online space. He argued that this connection itself justifies the city’s involvement. The cause of the failure of ad review remains unclear. One suggestion is that image screening at submission may have been too lax. Another possibility is that the risks after video conversion could not be assessed. The failure to detect reposts of identical ads is also an issue. It is unclear whether review relied on human labor or automated processing. Some view the transparency of the ad library as insufficient. Disclosure of advertiser information and payment routes will be key.

Platform Responsibility and Future Focus

Meta has maintained a zero-tolerance policy toward child exploitation. Even so, harmful ads passed through screening over a long period. The gap between its promise of pre-screening and the results drew criticism. The expansion of its advertising business can be seen as having increased the review burden. The misuse of generative AI appears to have exceeded what review standards anticipated. The boundary between still images and generated videos may have made judgments harder. Identifying victims and supporting image removal will also be challenges. The scale of reach within the European Union forces an international response. Distribution to the United States, Australia, and India also requires cross-border action. Meta is also expanding its communication platforms. As Discord Launches Native App on Meta Quest shows, its penetration into daily life is advancing. The expansion of contact points can be seen as also expanding responsibility for safety. Stronger advertiser screening linked to payment suspension will be a focus. Frameworks for information sharing with outside researchers will also be questioned. Sanctions and audits in case of recurrence are likely to be debated.

Editorial Opinion

On short-term impact. The request for discussions with the City Attorney will increase pressure to disclose ad review practices. Measures against reposts of identical ads and faster post-hoc removal are expected to advance within three to six months. Emergency takedown procedures triggered by outside researchers are expected to be established. Regional display data in the ad library and disclosure of advertiser information are also expected to be debated. On the long-term view. Ad standards for generative AI apps are expected to be tightened by explicitly addressing video conversion and undressing uses. Balancing child protection with the advertising business will require introducing independent audits. With cross-border distribution as a given, information sharing with authorities in each country is expected to become routine. Restoring user trust will require a record of victim support and recurrence prevention. Questions from the editors. Whether the promise of pre-screening can withstand the reality of mass ad placements is at issue. Whether the jurisdictional claim will delay redress for cross-border harm is at issue. Whether the lack of transparency is hindering external verification is at issue. Whether safety or business speed should take priority is at issue.

References

• “San Francisco Orders Meta to Stop ‘Allowing’ AI Child Abuse Ads”, by Matt Burgess — Wired, 2026-09-09T21:15:27.000Z (ARR)
• Source URL: https://www.wired.com/story/san-francisco-orders-meta-to-stop-allowing-ai-child-abuse-ads/