San Francisco City Attorney David Chiu didn’t mince words. He sent Meta a cease-and-desist letter this week. The demand was clear. Stop allowing paid advertisements featuring AI-generated child sexual abuse material on Facebook and Instagram. Explain how such content slipped through. Again and again.
The letter, first reported by WIRED, marks the latest escalation in a growing scandal. Researchers had already uncovered troubling patterns. But the city’s intervention adds legal pressure from Meta’s backyard. And it highlights a stubborn problem the social media giant has yet to solve.
Disturbing Scale of the Ads
Just one day earlier, the Tech Transparency Project laid out the evidence in stark detail. Its September 8 report identified more than 350 ads containing child sexual abuse material that ran across Meta’s platforms in recent months. Some transformed ordinary photos of minors into short videos depicting sexual acts. Others used still images of real children, including a young member of a European royal family and a 14-year-old social media influencer.
One ad took a press photo from the royal family’s official website. AI animation turned it into a video of the child performing a graphic sex act. Another pulled an image of a preteen girl in a pink athletic outfit with pigtails. The resulting clip showed her appearing frightened as an adult male molested her. A third featured the 14-year-old influencer in her sports club uniform. The video depicted her performing oral sex. These weren’t abstract creations. They involved identifiable real victims.
The vast majority — 298 out of 332 ads examined in one sample — promoted AI image and video editing apps. Many came from Chinese developers. They pitched “nudify” tools. Users could upload photos and generate explicit versions. Some ads carried voiceovers like “There are no restrictions. All the characters are real people and there are many options. This is the AI men actually use.” The ads reached over 29,000 accounts, including thousands in Europe, the U.S., India, and Australia. Total ad spend stayed low. Under $5,000 across the flagged campaigns. Impressions often numbered below 200 per ad. But the persistence mattered more than the numbers.
Katie Paul, director of the Tech Transparency Project, didn’t hold back. “What we’re talking about here is not simply AI-generated child sex abuse material,” she told reporters. The group reported the ads to Meta and the National Center for Missing and Exploited Children. Some came down. Yet new ones featuring the same children appeared even after notifications. “Meta has claimed it takes a hard stance against child sexual abuse material, but since TTP’s September 8 report published, more brand-new ads have run featuring some of the same children TTP sent to Meta,” Paul added.
This wasn’t the first warning. In August, WIRED revealed an earlier batch of over 50 such ads. Those ran between November 2025 and early August 2026. Meta removed them after being contacted. The company pointed to new AI detection tools rolled out recently. It insisted most had limited reach and many were blocked before wider distribution. “Sexual exploitation is horrific, and we work aggressively to keep it off our platform,” a Meta spokesperson said at the time.
But the follow-up investigation showed the tools fell short. Identical ads reappeared. Moderation proved inconsistent. Some ads stayed up for days after reports. Others were rejected initially only to run later under slight variations. The Tech Transparency Project criticized the company’s slow responses and policy application. In one instance, Meta took a week or more to act on flagged content and sometimes claimed the material didn’t violate standards.
Chiu’s letter cuts through the corporate language. “Meta has failed to address a known issue of AI-generated child sexually explicit ads, while profiting from them,” he said in a statement to WIRED. “These ads are deeply disturbing and cause harm to real children. Meta claims to have zero tolerance for child exploitation, yet the scale and persistence of these issues suggest current efforts are woefully inadequate. This is unacceptable, and Meta must do better.”
The cease-and-desist demands specific answers. How did the ads bypass moderation systems? How does Meta escalate and report such content to NCMEC? What actions does the company take against the advertisers and repeat offenders? Chiu’s office reviewed many of the ads flagged by researchers. The letter leaves little room for vague assurances.
Meta pushed back on jurisdiction. The ads didn’t target San Francisco users, the company argued. Therefore, the city attorney lacked authority. A spokesperson reiterated the standard line to multiple outlets, including SFGATE. “We don’t tolerate nudify apps or any kind of child exploitation, whether real or AI-generated,” Tracy Clayton said. “These criminals constantly shift tactics to evade detection, which is why we keep strengthening our detection and enforcement.” Many flagged ads had already been removed, the company added. It was taking action against some of the promoted apps.
Yet the pattern repeats. A July BBC investigation found similar ads running in India. The Indian government ordered Meta to remove them. Still, 84 of the ads in the latest sample appeared in India. Seventy-eight ran after that directive. In Australia, where officials examine Meta’s compliance with underage social media bans, 120 ads surfaced. The U.S. saw 262. Senator Mark Warner had already sent a pointed letter to Mark Zuckerberg in August. He demanded explanations and corrective action after the initial disclosures. Meta said it would respond.
The issue ties into broader concerns about Meta’s reporting systems. Earlier this year, law enforcement officials told The Guardian that Meta’s AI-generated tips to investigators often amounted to low-quality “junk” data. Internet Crimes Against Children task forces reported being overwhelmed. Valuable time got wasted. Meta disputed those claims. It pointed to cooperation with authorities and praised its processes by the Department of Justice and NCMEC. Still, the gap between automated detection and real-world enforcement shows.
These ads don’t exist in isolation. They promote tools that anyone can use to create more material. The apps land in Apple and Google stores. Some have been pulled after scrutiny, like Lumin – AI Partner. Others remain. The business model thrives on volume and evasion. Advertisers shift accounts, tweak copy, test new images. Moderation systems trained on known patterns struggle against rapid iteration. And when the content involves real children’s photos scraped from public profiles or news sites, the harm compounds. Victims face renewed exposure. Families deal with fresh trauma.
But the San Francisco action stands out. A local prosecutor targeting one of the world’s largest tech firms headquartered nearby sends a message. It bypasses federal inaction on some fronts. It forces Meta to address specifics rather than general commitments. Whether it leads to meaningful changes in detection, advertiser vetting, or faster takedowns remains to be seen. The company has removed the bulk of the reported ads now. Its ad library shows violations for child sexual exploitation and adult nudity policies.
Even so, researchers continue to spot fresh examples. The cat-and-mouse game persists. Lawmakers in Washington and regulators abroad watch closely. Australia’s eSafety commissioner has taken interest. European authorities could follow given the reach there. And Chiu’s demands for transparency on reporting and escalation could set a precedent for other cities or states.
Meta built its empire on engagement and advertising scale. That same machinery now surfaces content that exploits the most vulnerable. The firm’s promises of zero tolerance clash with the evidence of repeated failures. Short-term fixes like post-report removals don’t address root weaknesses in pre-approval scanning or advertiser accountability. As AI tools grow more accessible, the volume of potential abuse material will only increase. Platforms must adapt faster. Or face mounting legal and public consequences.
The latest cease-and-desist won’t solve everything. It does, however, put Meta on notice. From its headquarters in Menlo Park, the company can no longer treat this as a distant moderation problem. Real children. Real ads. Real profits, however small. The questions from San Francisco demand real answers.