At least one judge seems to think so. A federal appeals court has upheld a lower court ruling that dismissed a possession charge against a man who created AI-generated child sexual abuse material, finding the charge couldn’t survive First Amendment scrutiny under existing Supreme Court precedent. This is a decision that legal experts on multiple sides of the issue say exposes a genuine crisis in how the law is keeping pace with the technology.
The case involves Anderegg, who was charged with producing and distributing AI-generated material, transferring it to a minor under 16, and possessing it. A federal district court dismissed only the possession charge, leaving the other three intact, meaning Anderegg still faces prosecution for producing and distributing the material and for transferring it to a minor. The Justice Department appealed the dismissal specifically, arguing the government should be able to prohibit possession of obscene virtual child sexual abuse material even when no actual child was involved in its creation. The Seventh Circuit sided with the lower court and rejected that argument.
The legal logic underneath this ruling traces back two decades. Generally, the government can’t criminalize private possession of obscene material inside someone’s home, but it can prohibit possession of child sexual abuse material specifically because producing it requires exploiting a real child victim. That exception is what allows CSAM laws to exist at all under the First Amendment. In 2002, though, the Supreme Court complicated that framework in Ashcroft v. Free Speech Coalition, striking down restrictions on sexually explicit images of fictional children created without any real minor involved, reasoning that the government’s compelling interest in protecting real children from exploitation doesn’t extend the same way when no real child exists in the material. The Seventh Circuit’s ruling this week is essentially that 2002 precedent applied to AI-generated imagery. It’s the same legal logic, but applied to a very different generation of technology.
Hans von Spakovsky of Advancing American Freedom told Fox News Digital the appeals court “had no choice but to uphold the dismissal under the First Amendment” given that binding precedent. But he also endorsed the concerns raised in a concurrence by Judges Lee and Kolar, who explicitly flagged that the Supreme Court may need to revisit its own 2002 reasoning given how far AI has advanced. “Unfortunately, AI-generated images are now virtually indistinguishable from the real thing,” von Spakovsky said, “and some of these images can be created based on real child sexual abuse material. Technology may have overtaken and overwhelmed the reasoning used in those prior decisions.”
That last point is where this ruling gets alarming rather than just legally technical. Lee and Kolar’s concurrence cited research finding hundreds of actual child sexual abuse material images embedded in a major dataset used to train image-generating AI models, meaning some “AI-generated” content may not be as disconnected from real victims as the 2002 precedent assumed. The judges stopped short of resolving that question here, noting the government hadn’t raised it in this specific case and the record was too thin to settle it. But the door they left open is significant: if AI models are, even partially, trained on real abuse material, the entire legal distinction between “virtual” and “real” CSAM starts to collapse, and with it, the constitutional reasoning this ruling depends on. Separately, prosecutors have warned that photorealistic AI-generated CSAM is already complicating investigators’ ability to identify real victims, since increasingly convincing fakes make it harder to tell which images depict actual abuse and which don’t.
Justice Clarence Thomas, notably, saw this exact collision coming more than 20 years ago. In his own opinion on the 2002 case, he warned that if technology ever advanced to the point where virtual depictions became indistinguishable from real abuse material, the government might well have a compelling interest in regulating even lawful speech to enforce laws against the unlawful kind. That warning reads less like a footnote today than like a prediction that’s already arrived.
This is a case where the law hasn’t caught up to what the technology can now do, and every legal actor involved (the dismissing court, the appeals panel, the concurring judges, and even the outside legal commentators weighing in) seems to agree on that much, even while disagreeing about what to do next. With as fast as technology keeps changing, it’s time for the law to catch up.