Photo by dp singh Bhullar on Pexels · Pexels License
Everyone agrees on the problem. AI tools that strip real people into fake nude images have been used to harass teenage girls, humiliate women, and wreck reputations at industrial scale. The question in front of the Eighth U.S. Circuit Court of Appeals is narrower and, in some ways, harder: what is a state allowed to break while trying to fix that?
On November 9, the Liberty Justice Center filed an amicus brief urging the appeals court to block enforcement of Minnesota's HF 1606, a first-of-its-kind law titled the Prohibition on Nudification Technology. The brief supports SpaceXAI, Elon Musk's AI company and the maker of Grok, which sued Minnesota in late July, days before the law was set to take effect on August 1. A federal district judge denied the company's request for a preliminary injunction in early September, but the Eighth Circuit reversed that decision in October, pausing enforcement while the appeal proceeds.
Here is what makes the Minnesota law unusual, and legally shaky. Most states, Minnesota included, already have nonconsensual intimate imagery laws that punish the people who create or spread abusive deepfakes. HF 1606 takes a different route: it puts liability on the technology company whose tool was used, with penalties of up to $500,000 per image, enforced by the state attorney general. The law contains no intent requirement and no knowledge requirement. Under its terms, a company could owe half a million dollars even if it banned deepfakes, built safeguards to stop them, and a single user slipped past. It could owe the fine even if the person depicted consented, or created the image of themselves, or never shared it with anyone.
The definition at the heart of the statute is where things get genuinely strange. The law targets realistic depictions of 'intimate parts,' defined to include the groin, inner thigh, buttocks, or breast. As legal analysts at Lawfare pointed out, a realistic AI image of someone in a bathing suit could qualify. The statute contains no exception for satire, a category of speech the Supreme Court has long treated as protected. The company's original complaint argued the law bans ordinary depictions of men without shirts and people in swimwear.
Minnesota's defense, in court filings, amounts to arguing that none of this is a problem because consenting adults can simply take real nude photos of themselves instead. That is a striking thing for a state to say. It concedes that the law reaches far beyond nonconsensual abuse and treats the breadth as a feature.
This matters well beyond one Musk company. When liability is strict and fines are half a million dollars per image, the rational corporate response is to shut features off entirely rather than risk enforcement. The Liberty Justice Center's brief makes exactly this point: providers could respond by cutting off image services in Minnesota, and the resulting chill would reach artists, small businesses promoting themselves, nonprofits, and ordinary people engaged in political argument. That is how overbroad laws actually work in practice. They rarely get tested in court by the powerful, who can afford lawyers. They get quietly obeyed by everyone else.
There is a longer history here worth remembering. Content moderation rules written broadly have a consistent track record of falling hardest on the people with the least institutional power. Platforms under legal pressure overcorrect, and overcorrection tends to hit Black creators, activists, and satirists whose work is disproportionately flagged and removed under automated enforcement. A legal regime that pressures AI companies to restrict image generation by default invites the same pattern at the creation stage, before a single image is even posted. The communities most often depicted in deepfake abuse deserve real protection. They also deserve laws that actually punish the abusers rather than constricting everyone's tools.
The Eighth Circuit has already signaled skepticism by halting the law in October, albeit in a one-sentence order with no reasoning. The appeal now forces the court to answer a question that will echo through every statehouse watching this case: can the government ban millions of lawful uses of a technology to get at a narrower category of harmful ones? The answer Minnesota needs is yes. The answer the First Amendment has historically given is no.
Free Game Takeaway
Watch who gets regulated, not just what gets regulated. Minnesota already had laws punishing people who make and spread abusive deepfakes; HF 1606 instead targets the toolmakers, which is why it threatens lawful speech while arguably doing less to stop determined abusers who can use offshore or open-source tools. If you create content, run a business, or use AI tools for work, cases like this one will determine what features stay available to you, because strict-liability laws push companies to remove capabilities rather than defend them. The signal to watch next is the Eighth Circuit's full ruling on the injunction appeal in SpaceXAI LLC v. Ellison, No. 26-2806: a decision striking down the law would discourage copycat strict-liability bills in other states, while a decision upholding it would likely trigger a wave of similar legislation and a rapid shrinking of AI image features nationwide.