A federal appeals court has now confronted one of the most disturbing consequences of artificial intelligence: computer-generated child pornography so realistic that it can be virtually indistinguishable from photographs of actual abuse. The result should trouble Americans across the political spectrum. In United States v. Anderegg, decided August 25, the U.S. Court of Appeals for the Seventh Circuit affirmed the dismissal of a federal charge involving the private possession of obscene, entirely AI-generated child sexual abuse material. The court did not broadly legalize the creation or distribution of such material. Rather, it concluded that existing Supreme Court precedent protects private possession in the home when no actual child was used to create the images.
That distinction matters legally. Morally and socially, however, it presents a deeply unsettling problem.
The defendant, Steven Anderegg, was accused of using generative AI to produce hyper-realistic sexual images depicting prepubescent children. Prosecutors charged him with multiple offenses, including production, distribution, possession, and transferring material to a minor. The ruling at issue concerned the possession count under 18 U.S.C. §1466A(b)(1). The Seventh Circuit concluded that Supreme Court precedents, particularly Stanley v. Georgia and Ashcroft v. Free Speech Coalition, prevented the government from criminalizing private possession of obscene virtual material when no real child was harmed in its production.
The judges themselves recognized the uncomfortable implications. Their essential position was not that AI-generated child pornography deserves societal approval, but that lower courts cannot simply redraw constitutional boundaries established by the Supreme Court because technology has made those boundaries increasingly difficult to defend.
That is an important distinction, but it does not make the consequences less concerning.
Conservatives have traditionally defended both constitutional limits on government and society’s authority to protect children. Those principles need not be contradictory. The First Amendment is indispensable precisely because unpopular and offensive speech requires protection from governmental censorship. Yet obscenity has never enjoyed unlimited constitutional protection, and child pornography involving actual children is categorically outside First Amendment protection because its creation records and perpetuates genuine abuse.
Artificial intelligence has now created a category that the architects of those doctrines could scarcely have imagined.
The traditional argument for protecting purely virtual depictions rests heavily upon the absence of a victim in the production process. No child posed for the image; therefore, no child was abused to manufacture it. That reasoning has considerable constitutional force. But generative AI changes the practical question. A person can now deliberately instruct a machine to manufacture photorealistic depictions of crimes against children on an unprecedented scale. The resulting material can potentially circulate alongside genuine abuse imagery, complicate investigations, normalize increasingly extreme material within predatory communities, and create a marketplace or subculture in which the distinction between simulated and actual exploitation becomes increasingly difficult to police.
The law cannot responsibly pretend that this transformation is merely the digital equivalent of an obscene drawing locked inside someone’s desk.
There is also an important limitation to the decision that should not be lost amid sensational headlines: it does not establish a general constitutional right to produce and distribute AI-generated child pornography. Indeed, federal prosecutions involving AI-generated obscene material continue. Earlier this month, for example, a Kansas defendant pleaded guilty to distributing obscene computer-generated images depicting minors engaged in sexually explicit conduct. In February, federal prosecutors secured a conviction in Ohio involving both actual CSAM and obscene AI-generated depictions after a court rejected a broader First Amendment challenge.
The real question, therefore, is narrower and more difficult: Should the Constitution prevent government from punishing an adult solely for privately possessing entirely synthetic, obscene sexual depictions of children?
There are legitimate civil-liberties reasons for approaching that question carefully. Giving the government unlimited authority to criminalize possession of imaginary images could establish precedents extending far beyond this uniquely repellent subject. Constitutional protections are tested most severely when the material being protected is something decent people despise.
But constitutional restraint should not become constitutional paralysis.
Congress should examine the statutory framework immediately, carefully distinguishing possession, production, distribution, solicitation, transmission to minors, and AI imagery based upon identifiable real children. Legislators should enact laws that address demonstrable harms caused by synthetic CSAM rather than relying on statutes designed for an earlier technological era.
The deeper lesson is uncomfortable. Artificial intelligence has exposed a gap between legal doctrine and technological reality. A machine can now generate in seconds material that previously required either an artist’s imagination or the horrific exploitation of an actual child.
Protecting constitutional liberty requires precision. Protecting children requires resolve. America should be capable of doing both without accepting the proposition that technological novelty leaves society powerless against conduct that deliberately simulates, normalizes, and potentially facilitates the sexual exploitation of children.