Key Takeaways

  • The Seventh Circuit upheld the dismissal of a possession charge involving AI-generated abuse imagery that did not depict real children.
  • The defendant still faces charges tied to production, distribution, and sending images to a minor.
  • AI providers and online platforms face growing pressure to detect synthetic CSAM even where possession laws remain unsettled.

A federal appeals court has upheld the dismissal of a criminal possession charge against a 42-year-old Wisconsin man accused of creating thousands of explicit, AI-generated images depicting children.

The Seventh Circuit concluded that private possession of synthetic child sexual abuse material, when no real child is depicted, is protected by the First Amendment under existing Supreme Court precedent. The decision does not provide comparable protection for producing or distributing the material, or for transmitting it to minors.

Prosecutors say the defendant created thousands of “photorealistic” images and sent some of them to a 15-year-old boy, along with an explanation of how he produced them using AI. He still faces charges covering production, distribution, and transmission to a minor. The dismissal applies only to possession.

According to WEAU, a federal judge in Wisconsin initially dismissed the charge before the Seventh Circuit upheld that decision. The judge cited Supreme Court precedents from 1969 and 2002 addressing the private possession of obscene material and virtual imagery that does not involve an actual child.

“Given the relentless advancement in artificial intelligence... we have some concerns about the lines these cases draw, but we are not free to redraw them ourselves,” the judge wrote.

The decision is narrower than a declaration that all AI-generated abuse imagery is lawful. Production and distribution can still expose defendants to prosecution, particularly when the material is sent to a child or used in grooming, harassment, extortion, or another offense. But the ruling creates a difficult boundary for investigators who discover synthetic material stored privately and cannot establish that a real minor was depicted.

For technology businesses, that boundary does not eliminate trust and safety obligations. A platform can prohibit synthetic CSAM through its terms of service even when private possession falls outside a particular criminal statute. Hosting providers, messaging services, generative-AI developers, and enterprise collaboration vendors also operate across jurisdictions with different laws.

Detection systems must now distinguish synthetic content from imagery involving a real victim when both can appear photorealistic. Automated classifiers may identify suspected abuse content, but determining whether an image is synthetic can require additional technical and investigative work. Metadata may be missing or manipulated. Watermarks can be removed. Models can also be adapted or operated locally, outside the controls imposed by commercial AI services.

Reports to the National Center for Missing and Exploited Children involving generative-AI child abuse content increased 1,325%, from 4,700 in 2023 to more than 67,000 in 2024. Separately, the Internet Watch Foundation recorded a 380% increase in the AI-generated CSAM it actioned between 2023 and 2024, with 98% of that material depicting girls.

AlgorithmWatch reported in August 2026 that Internet Watch Foundation analysts identified more than 260 times as many AI-generated CSAM videos in 2025 as in 2024. Synthetic material accounts for around 1% of all identified CSAM, but its growth trajectory is steep (source).

Governments are beginning to address creation tools rather than relying only on possession and distribution statutes. The UK government announced legislation in November 2025 intended to target AI child abuse imagery at its source. The Seventh Circuit ruling illustrates why that approach is gaining attention: laws written around photographic evidence of abuse can encounter constitutional and evidentiary complications when generative systems create fictional but highly realistic subjects.

A sexual assault program director warned that synthetic imagery can still be used to blackmail and extort children. “People who are going to hurt another person will use whatever method that they can,” the director said.

Companies deploying generative models may respond by tightening prompt controls, monitoring abuse patterns, preserving auditable safety records, and establishing escalation procedures for reports involving minors. Platforms also need policies that separate criminal-law thresholds from their own content standards. The court may have drawn a constitutional line around private possession, but businesses still have to decide what they will generate, store, transmit, and report.