Image credit: Suzanne Kantra/Techlicious generated by ChatGPT
A federal appeals court has ruled that privately owning AI-generated images of child sexual abuse, made without real child involvement, is protected under the First Amendment if the images stay in your own home. The Seventh Circuit affirmed a lower court's decision to dismiss a possession charge against Steven Anderegg, who used the AI image generator Stable Diffusion to create hundreds of photorealistic images depicting minors in sexual situations.
Anderegg was caught when he sent an image to a minor over Instagram. Meta flagged that Instagram message to the National Center for Missing and Exploited Children, which led investigators to a search warrant and hundreds of AI-generated images on Anderegg's devices. He was charged with four offenses: producing, distributing, and possessing the images, plus transferring one to a minor. The appeals court ruling addresses only the possession charge.
The court's reasoning rests on two prior Supreme Court cases. In Stanley v. Georgia (1969), the court ruled that the government can't prosecute someone for privately owning obscene material inside their own home. And in Ashcroft v. Free Speech Coalition (2002), the court decided images that only resemble child sexual abuse material are protected speech, because the usual reasons for banning such material – that it fuels demand for real abuse, grooms children, or normalizes the behavior – don't apply when no real child exists. "The Government may not suppress lawful speech as the means to suppress unlawful speech," the court wrote, quoting Free Speech Coalition. "Protected speech does not become unprotected merely because it resembles the latter."
Judge John Lee, writing for the appeals panel, said the government's arguments that the images could groom children, normalize abuse, substitute for real abuse material, or make prosecutions harder were all "expressly discussed and rejected" by the Supreme Court in Free Speech Coalition. The panel said it remains bound by that precedent regardless of how far AI has advanced since 2002, warning that ignoring it would invite "anarchy...to prevail within the federal judicial system."
Lee and Judge Joshua Kolar, two of the three judges who ruled for Anderegg, also wrote separately to say the Supreme Court needs to revisit Free Speech Coalition itself. Today's AI-generated images, they argued, are "virtually indistinguishable from CSAM created through the physical abuse of a child," a possibility they say Justices Sandra Day O'Connor and Clarence Thomas anticipated in 2002 when they suggested the government might someday have a stronger case if the technology caught up.
The concurrence also raised something Free Speech Coalition never had to weigh: some AI image generators may have been trained on real child sexual abuse material, which would undercut the "no real victim" premise the ruling depended on. The judges said the record in this case doesn't allow them to evaluate that question, only to flag it for a higher court.
This ruling covers one circuit, Illinois, Indiana, and Wisconsin, and one narrow charge. It doesn't settle the question nationally, and two of the judges are asking the Supreme Court to take a harder look. As AI image generators keep improving, expect more cases like this one to work through the courts before there's a settled, nationwide answer.
