Seventh Circuit Holds Private Home Possession of AI-Generated CSAM Protected Under Stanley and Ashcroft, While Production and Distribution Remain Prosecutable

A three-judge panel of the Seventh Circuit held on August 25, 2026 in United States v. Anderegg, No. 25-1354, that 18 U.S.C. § 1466A(b)(1) is unconstitutional as applied to a defendant charged with possessing AI-generated child sexual abuse material in his home, because the images “do not depict an actual child.” Judge John Z. Lee wrote for the panel, joined by Judges Doris Pryor and Joshua Kolar; Lee and Kolar also filed a concurrence.

The holding is far narrower than its coverage. The court said so itself: “The appeal in this case deals only with Anderegg’s challenge to the home possession charge, not to the production or distribution charge.” Steven Anderegg was arrested in 2024 on charges of fabricating, possessing, and distributing AI-generated CSAM and sending some of it to a minor. Production, distribution, and transfer to a minor all remain prosecutable. What the ruling protects is private possession, by a specific defendant, of images depicting no real child.

Two Supreme Court precedents produced that result: Stanley v. Georgia (1969), holding that private possession of obscenity in the home is protected, and Ashcroft v. Free Speech Coalition (2002), holding that material not depicting real children is not child pornography and is protected speech. The intersection is the gap — private possession plus no real child — and the panel found no room to rule otherwise.

The judges said plainly that they thought the precedent was outrun by the technology, and that they could not fix it:

“In Free Speech Coalition, the Supreme Court addressed the scope of First Amendment protections for virtual CSAM, but that was nearly twenty-five years ago, and the image-generation technology available today was likely unimaginable back then. 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.”

That last clause is why this belongs in a capture timeline. Ashcroft’s reasoning rested on a factual premise — that virtual images are distinguishable from real ones and that no child is harmed in making them. Generative models did not overturn that premise by argument; they eroded the conditions that made it operative, at a speed no appellate court can match. The doctrine did not change. What the doctrine describes did.

This is the inverse of the AI-governance pattern this corpus otherwise records, where federal actors move to clear state law out of the way of AI development — the EO 14365 litigation task force, its first deployment against Colorado’s AI Act, the Great American AI Act’s three-year preemption. Here no one is clearing anything. A court applied settled law faithfully and produced an outcome it flagged as a problem, then said the remedy is not judicial. The gap is left for the Supreme Court or Congress, and the panel’s concurrence is effectively an invitation to one of them.

Sourcing note. The case name, docket number 25-1354, August 25 date, panel composition, the as-applied holding on § 1466A(b)(1), the explicit limitation to the home-possession count, and the quoted passages are from the Volokh Conspiracy’s account at Reason and corroborated by Snopes’s fact-check, with Minnesota Lawyer and The Daily Record covering the same ruling. The opinion itself was not read here. Note that the capture ledger’s fact-check flagged — correctly and usefully — that it could not confirm which Supreme Court precedent controlled and therefore excluded Ashcroft from its summary rather than supplying it from memory; both precedents are named here because sources were found that state them. Headline framings in wider coverage that this ruling makes AI-generated CSAM broadly legal are contradicted by the opinion’s own scope and should not be repeated; the district court had already dismissed the possession count below, and this affirms that narrow result.