A federal appeals court just ruled that the Constitution protects the private possession of AI-generated child sexual abuse material, and the judges who wrote the ruling are furious about it. The Seventh Circuit didn't want to issue this ruling. They said so, explicitly, in the opinion. But they say the Supreme Court tied their hands with a 2002 decision written when the most sophisticated image-generation technology on earth was a Motorola Razr.
What the Court Actually Said
Steven Anderegg, a Wisconsin man, was found with hundreds of AI-generated sexually explicit images depicting children on devices seized from his home. He faced four charges: producing the material, distributing it, transferring it to a minor under 16, and possessing it. A federal district court threw out the possession charge on First Amendment grounds. The Justice Department appealed. They lost.
The Seventh Circuit, as Fox News reports, said it was bound by the Supreme Court's 2002 decision in Free Speech Coalition v. Ashcroft, which struck down restrictions on sexually explicit depictions of fictional children when no actual child was involved in creating them. The reasoning back then was that the government's interest in protecting real children from exploitation doesn't apply when no real child was exploited. In 2002, that distinction made some legal sense. In 2026, when AI can generate photorealistic images of child abuse from a text prompt, it has become a disaster.
The three other charges against Anderegg, including distributing the material and transferring it to a minor, remain intact. This ruling is specifically about whether the government can criminalize simply having this material in your home. The answer, for now, is no.
The Judges Know This Is Wrong and Said So Anyway
This is the part of the story that deserves to be read slowly. Judges John Lee and Joshua Kolar, who concurred with the ruling, did not quietly sign on and go home. They wrote separately to flag that the precedent they were forced to apply is catastrophically out of date.
"Given the relentless advancement in artificial intelligence models, we have some concerns about the lines these cases draw, but we are not free to redraw them ourselves," Judge Lee wrote, according to Fox News. That is a federal judge, in an official legal opinion, essentially writing a distress flare addressed to the Supreme Court.
The concurrence also raised something genuinely alarming: research has found hundreds of real child sexual abuse images embedded in major datasets used to train image-generating AI. If the models themselves were trained on material depicting the abuse of actual children, the legal fiction of a clean separation between "fictional" AI images and real-world exploitation starts to collapse entirely. The judges acknowledged the record in this case wasn't sufficient to resolve that question, but they made sure to put it on paper.
Clarence Thomas Called This Shot in 2002
Here is an uncomfortable sentence to type: Clarence Thomas saw this coming. In his concurrence to the original Free Speech Coalition ruling, Thomas warned that advancing technology could eventually make virtual depictions of child abuse indistinguishable from material depicting actual victims, which would gut the legal reasoning behind the decision before the ink was dry.
That warning sat in a footnote for twenty-five years while the technology caught up. It has now fully caught up. The Justice Department has separately warned, as Fox News notes, that the explosion of AI-generated material is already bogging down investigators trying to identify real abuse victims, because increasingly realistic synthetic images are flooding the same channels where genuine evidence of crimes against children would appear. Investigators have to examine all of it. Every. Single. Image.
So the practical consequence of the current legal framework is not some abstract First Amendment principle at work. It is investigators drowning in synthetic material while real victims wait.
Where the Law Actually Stands Right Now
The legal architecture here is genuinely complex, so here is what is and is not currently allowed. The government can criminalize the production, distribution, and transfer of AI-generated CSAM. It can absolutely prosecute possession of material depicting real children. What it cannot currently do, per this ruling, is prosecute someone purely for having AI-generated images on their home devices, because the Supreme Court's 2002 precedent carved out private possession of obscene material not involving real victims as constitutionally protected.
Congress has tried to patch some of this. Melania Trump's Take It Down Act, which Fox News reports just secured its first conviction, targets the non-consensual sharing of AI-generated explicit images, but that law is primarily aimed at deepfake revenge porn targeting adults and minors in distribution contexts, not the possession question the Seventh Circuit just addressed.
The gap this ruling exposes is real, specific, and the Supreme Court is the only institution that can close it. Whether this case or a case like it reaches the justices, and how quickly, is now the central question.
The Dingo Take
You are supposed to read this ruling and feel reassured by the fact that three other charges are still standing, that the system is still working, that this is just a narrow procedural outcome about one specific charge. Do not feel reassured. What happened here is that a federal appeals court was legally compelled to protect the in-home possession of AI-generated images of child sexual abuse because nine justices in 2002 made a decision about technology they could not have imagined, and no one with the power to update that decision has done so in the twenty-five years since.
The Seventh Circuit basically wrote a cover letter to the Supreme Court begging them to take this up and fix it. The Justice Department is watching AI-generated material swamp investigations into real crimes against real children. Congress has passed piecemeal bills that don't address the core gap. And the Supreme Court, which has shown no shortage of willingness to overturn precedent when it wants to, has not moved on this. At some point "we were bound by prior precedent" stops being an explanation and starts being an excuse.
Anderegg still faces three serious charges and the legal system may yet put him away for a long time. But the next person, and the person after that, can possess AI-generated child abuse material in their home and the federal government currently cannot touch them for it. That is the world this ruling describes. Someone should probably be in more of a hurry to change it.
Comments