Making and sharing AI-generated child sexual abuse material is a crime in both the US and UK. A 2026 appeals ruling muddied one narrow corner of that, and the headlines lost the nuance. Here is the accurate picture.
The question of whether AI-generated CSAM is illegal used to have a short answer. Then, in August 2026, a federal appeals court handed down a ruling that spawned a wave of alarming headlines suggesting the courts had legalized it. That is not what happened, but the truth is more complicated than either the panic or the reassurances suggest.
So let’s walk through what the law actually says, where the new ruling fits, and what is still very much a crime. This is a legal explainer, not legal advice, and the distinctions here are precise for a reason.
At a Glance
Making, sending, or receiving AI-generated child sexual abuse material is a federal crime in the United States, and any image that depicts a real child is illegal outright. The one unsettled piece, after an August 2026 appeals ruling, is whether privately possessing fully synthetic images with no real victim is protected speech. Producing or sharing them is not.
Table of Contents
The Short Answer, With One Big Asterisk
In the United States, generating, distributing, or receiving sexual images of children is illegal whether a camera or an algorithm made them. That has not changed. The asterisk is narrow: one appeals court has said that quietly possessing purely synthetic images, ones that depict no real child and are never shared, may be protected by the First Amendment. Everything around that act, the creating and the sending, stays criminal.
Two ideas do most of the work in this area of law, so it helps to separate them: whether an image involves a real child, and whether the conduct is possession or distribution. The legal answers change depending on where a case lands on those two axes.
What US Federal Law Actually Says
Federal law splits into two tracks. Material depicting a real child, including a real child’s face edited onto explicit content, is prosecuted under standard child-exploitation statutes. There is no gray area there. A deepfake of an actual minor is treated as abuse material, full stop.
Wholly synthetic images, with no real child anywhere in the training or the output, run through a different statute aimed at obscene visual depictions of minors, often cited as Section 1466A. That law exists because of a 2002 Supreme Court decision, Ashcroft v. Free Speech Coalition, which held that the government cannot ban virtual depictions simply for existing, but left room to prohibit material that meets the legal test for obscenity. Congress responded with the PROTECT Act, and Section 1466A is the piece that still applies to computer-generated content today.
Separately, 46 states had enacted their own laws criminalizing AI-generated or computer-edited abuse material as of August 2026, according to a running tally from Enough Abuse. So even where a federal charge gets complicated, a state charge often does not.
The 2026 Ruling That Changed the Conversation
The headlines came from United States v. Anderegg. The defendant used an image generator to produce photorealistic sexual images of children and was charged with producing them, distributing them, possessing them, and sending obscene material to a minor.
On August 25, 2026, the Seventh Circuit Court of Appeals let stand the dismissal of just one of those charges: private possession. The court’s logic leaned on a 1969 case, Stanley v. Georgia, which protects the right to possess even obscene material in the privacy of your own home. Because the images depicted no real child, the court reasoned, the usual exception that strips that protection for abuse material did not apply. Legal scholar Eugene Volokh walked through the reasoning shortly after.
Here is the part the scary headlines skipped. The court did not strike down the statute. It did not create a right to produce these images, to distribute them, to send or receive them, or to possess anything involving a real child. Three of the four charges stood. The ruling binds only the Seventh Circuit, and a split among the circuits is exactly the kind of thing that eventually reaches the Supreme Court. Treat it as one narrow crack, not an open door.
The States and the United Kingdom
Outside that single federal carve-out, the law has been moving hard in the opposite direction.
The United Kingdom went furthest. Its Crime and Policing Act 2026 makes it a crime to make, possess, or share AI-generated abuse imagery, and it goes a step past most US law by criminalizing the AI tools built specifically to generate it. Possessing one of those generators can carry up to five years, and platform operators who facilitate the abuse face up to ten. There is no private-possession exception in that framework.
Back in the US, the state patchwork keeps thickening, and lawmakers have floated federal updates aimed squarely at closing the synthetic-content gap the Anderegg case exposed. The direction of travel among legislatures is clear even where one court paused.
Why This Matters Beyond a Single Case
Most people will never touch a courtroom over this, so why does the fine print matter?
Because it shapes enforcement everywhere else. Platforms do not wait for constitutional questions to resolve. X, for instance, treats AI-generated and real CSAM identically and reports both to authorities, as we covered in our breakdown of how X enforces its zero-tolerance policy. The fine print also matters for the developers building image models, who face growing liability for tools that can be pushed to generate this content, and for the real survivors whose likenesses get pulled into deepfakes, which the law protects without ambiguity.
This article is general information, not legal advice. Laws on AI-generated abuse material vary by country and state, are changing quickly, and are being actively litigated. Consult a qualified lawyer or your local authorities for guidance on a specific situation.
Frequently Asked Questions
Is it illegal to create AI-generated CSAM?
Yes. Producing it is a crime under US federal law and in the UK, and the August 2026 appeals ruling did not change that. That ruling touched only private possession of fully synthetic images, not their creation.
Did the court legalize AI-generated CSAM?
No. The Seventh Circuit allowed one possession charge to be dismissed for images depicting no real child. Producing, distributing, sending, and receiving such material remained illegal, as did anything involving a real child. It applies only in that circuit.
Is making a deepfake of a real child illegal?
Yes, unambiguously. When AI edits or depicts an actual child, it is prosecuted as child sexual abuse material under standard federal statutes, with no synthetic-content loophole.
What does UK law say?
The UK’s Crime and Policing Act 2026 bans making, possessing, and sharing AI-generated abuse imagery and criminalizes the AI tools built to produce it. Penalties reach five years for possessing such a generator and ten for platform operators who enable the abuse.
Can AI companies be held liable for what their tools generate?
Increasingly, yes. Lawmakers and regulators are targeting developers whose models can be prompted into producing this content, and the UK law explicitly reaches tools optimized for it. Expect liability questions to sharpen, not soften.
Closing Thoughts
The honest answer to “is AI-generated CSAM illegal” is yes, with a single asterisk that one court placed on private possession of purely synthetic images. Read past the headlines and the law is fairly firm: creating this material and sharing it is criminal, anything involving a real child is criminal, and the UK has gone further still. The Anderegg ruling is a live legal question, not a green light, and it may not survive higher review. For more on how platforms and regulators handle these questions, browse YouGottaRead’s Tech section.
