US AI regulation

xAI's Minnesota Nudification Lawsuit Exposes a Real Drafting Flaw in an Otherwise Justified Law

Minnesota's HF 1606 nudification ban took effect Aug. 1 after a judge denied xAI's injunction bid, but xAI's overbreadth and strict-liability claims have merit.

Minnesota's Nudification Ban, By the Numbers People of Internet Research · US $500,000 Minimum civil penalty per violation Applies per unlawful access, downl… 132-1 / 65-0 Legislative vote margin HF 1606 passed the Minnesota House… 5 days Days between filing and effective date xAI sued July 27 and sought emerge… 48 hours Federal takedown window, by contrast The TAKE IT DOWN Act requires plat… peopleofinternet.com
Minnesota's Nudification Ban, By the N… People of Internet Research · US $500,000 Minimum civil penalty per viol… 132-1 / 65-0 Legislative vote margin 5 days Days between filing and effec… 48 hours Federal takedown window, by contr… peopleofinternet.com

Key Takeaways

A first-of-its-kind ban survives its first test

Minnesota's House File 1606 — the nation's first dedicated ban on AI "nudification" services — took effect on schedule August 1, 2026, after U.S. District Judge Donovan W. Frank declined to block it. xAI, which sued the state on July 27 in the U.S. District Court for the District of Minnesota (Case No. 0:26-cv-03425), had asked for emergency relief on July 29, just three days before the law's effective date and roughly three months after Governor Tim Walz signed it. Frank denied the temporary restraining order primarily on timing grounds, writing that "such a delay in bringing the action and the motion suggests that harm is not immediate." The court converted xAI's request into a preliminary injunction motion, with a hearing set for August 19 in St. Paul.

The law passed the Minnesota House 132-1 and the Senate 65-0 — about as close to political consensus as a legislature gets. That near-unanimity reflects a real and legitimate harm: AI tools that strip clothing from a real person's photo without consent are being used at scale against classmates, ex-partners, and public figures, and the resulting images cause the kind of humiliation and reputational damage that ordinary defamation or privacy law wasn't built to move fast enough to stop. Minnesota Attorney General Keith Ellison summed up the case for the law bluntly: "AI nudification robs the target of their dignity... There are plenty of worthy debates to have about AI policy. This is not one of them." That framing is fair. Nonconsensual sexualized deepfakes are not a gray area, and a state legislature responding to them with urgency is doing its job.

Where the statute overshoots

The problem is not that Minnesota banned nudification. It's how it wrote the ban. HF 1606 imposes a civil penalty of not less than $500,000 per unlawful access, download, or use against anyone who "owns or controls" a website, app, or service that lets a user generate such an image — with no requirement that the operator knew about the misuse, intended it, or failed to take reasonable precautions. xAI's complaint, filed against Ellison in his official capacity, does not dispute Minnesota's interest in banning nonconsensual nude deepfakes; it argues the statute "extends far beyond that goal." That is a defensible reading of the text.

Two drafting choices do the damage. First, the law defines "intimate part" by importing the definition from Minnesota's criminal sexual conduct code (Minn. Stat. § 609.341, subd. 5) — genital area, groin, inner thigh, buttocks, or breast — rather than the narrower genitals-and-nipples definition used in Minnesota's own pre-existing deepfake statute, which the legislature considered and rejected in favor of the broader one. xAI argues that sweep is broad enough to cover ordinary swimwear and shirtless-torso imagery with no sexual content at all, and nothing in the statute's text distinguishes a nonconsensual sexual deepfake from a satirical political image or a shirtless beach photo altered for a meme. Second, the strict-liability structure attaches the moment a user succeeds in generating a prohibited image, regardless of what filters, moderation, or takedown systems the platform had in place — a standard that punishes companies for their users' conduct even when they built reasonable safeguards. xAI's complaint also notes the law exempts imagery produced by a human using "significant technical skill" (traditional photo editors, for instance) while punishing functionally identical output from an AI tool — an odd place to draw a free-speech line if the harm being addressed is the image itself, not the software used to make it.

A federal model already exists that threads this needle

Congress did not need strict liability to address this problem. The federal TAKE IT DOWN Act, signed into law May 19, 2025, requires covered platforms to remove nonconsensual intimate images — including AI-generated ones — within 48 hours of a valid notice, with FTC enforcement of the notice-and-removal duty beginning May 19, 2026. Crucially, that statute does not impose strict liability on platforms for a user's underlying act; it requires action on notice and grants a good-faith safe harbor for compliant takedowns. That is a narrower, notice-based model that still gives victims fast recourse without exposing every AI image tool operating in Minnesota to $500,000-per-instance liability for content they neither created nor could reasonably have prevented.

Minnesota's private right of action — allowing victims to seek compensatory damages (up to treble for mental anguish), punitive damages, injunctive relief, and attorney's fees — is a sound complement to AG enforcement and should survive any narrowing of the statute. The $500,000 civil penalty floor is not, on its own, the constitutional problem; deterrence needs teeth. The problem is applying that penalty without any knowledge requirement to an overbroad definition of the conduct it punishes.

What should happen next

Judge Frank's TRO denial was about timing, not the merits — xAI waited too long to ask for emergency relief, not that its underlying claims are weak. The August 19 preliminary injunction hearing is where the actual First Amendment and overbreadth arguments get tested, and on the current record, xAI has identified a genuine defect: a law can target nonconsensual sexual deepfakes precisely, the way its own prior deepfake statute did, without also sweeping in swimwear photos and satire under a strict-liability sledgehammer. Minnesota does not need to choose between protecting victims and drawing careful lines — it already had a narrower definition on the books before this bill widened it. The fix is not to strike the law down wholesale but for the legislature, or the court on a narrower injunction, to push the definition back toward actual nudity and add a scienter or safe-harbor standard for platforms that act on notice. That would preserve the penalty structure victims need while removing the parts of the statute that make it vulnerable — and that, more than the outcome of one TRO, is what the next few weeks in St. Paul should be about.

Sources & Citations

  1. xAI complaint, X.AI LLC v. Ellison (D. Minn., filed July 27, 2026)
  2. Minnesota Legislature — HF 1606 bill status (Chapter 72)
  3. Valley News Live — Judge denies xAI bid to block Minnesota AI nudification ban
  4. CBS News Minnesota — xAI sues to stop Minnesota nudification law
  5. Reclaim The Net — xAI's First Amendment challenge to Minnesota's nudification law
  6. Skadden — The TAKE IT DOWN Act notice-and-takedown summary