DOJ Backs xAI Challenge to Minnesota Nudify Ban

On August 19 in St. Paul, Judge Donovan Frank heard two hours of argument over whether Minnesota may keep blocking AI "nudification" tools while xAI's First Amendment suit proceeds. The day before, the Justice Department filed a statement of interest that did not ask for an injunction and still managed to stand with Elon Musk's company against the state's statute.

Minnesota's HF 1606 took effect August 1. It forbids anyone who owns or controls a website, app, or software from letting users sexually alter images with AI, with civil penalties that can reach $500,000 per violation. Liability is strict: platforms can be on the hook whether or not they meant the result. xAI sued in late July, arguing the law treats a general-purpose image model like a dedicated abuse service.

6 min read

In court, xAI attorney Robert Dunn said the legislature left no carve-outs for consent, art, parody, or a person altering their own photo. Minnesota Assistant Attorney General Janine Kimble answered that terms of service arrive after the fact, once someone has already seen, reported, and located an image. The state cited research claiming roughly 95 percent of deepfake imagery is nonconsensual and sexually explicit.

What Washington filed

Associate Attorney General Stanley Woodward signed the Justice Department's brief. It says the federal government shares Minnesota's goal of fighting sexual abuse imagery, then objects to how the state wrote the rule. Federal statutes such as the PROTECT Act and the TAKE IT DOWN Act, the filing argues, require knowledge, carve out matters of public concern, and center adult nonconsent. HF 1606 has none of those limits, and can reach an AI image of a shirtless adult in a pool because Minnesota borrows a broad definition of "intimate part."

The same brief ties the case to the Trump administration's push for a single national AI standard rather than "50 discordant state ones." "States must be careful not to impose excessive measures that hinder American national and economic security," the government wrote. That sentence does more work than the First Amendment footnotes. Minnesota may have drafted an overbroad statute. Washington is also auditioning a preemption theory that treats state AI rules as a national-security drag.

Judge Frank already denied xAI a temporary restraining order on July 31, noting the company waited until days before the August 1 effective date, nearly three months after the bill was signed. On Wednesday he took the preliminary injunction under advisement. For now the law still binds xAI. Europe is moving on nudify apps from another direction, and Sonar has watched Brussels force labels onto authentic-looking deepfakes. Minnesota chose a ban on the tool instead of a transparency sticker.

The load-bearing philosophical fight is whether Grok Imagine is a speaker or a machine that enables speakers. Minnesota wants the court to treat the product as a controllable instrument. xAI wants the court to treat generative edits as protected expression that a state cannot police with strict liability. Both frames can be half true. A general image model can produce satire, self-portraiture, and nonconsensual abuse with the same sampler. A legislature that refuses to separate those uses invites a constitutional wreck. A federal government that answers every state experiment with "wait for us" invites a vacuum where the abuse already lives.

Frank's forthcoming order will decide whether Minnesota can keep the gate closed while the merits crawl. Whatever he writes, the Justice Department has already said the quieter part out loud: the administration wants fewer state AI codes, and it will show up in district court to say so. Dignity interests are real. So is the risk that "national standard" becomes a euphemism for no standard until Congress finishes arguing about one.