First Amendment or Image-Based Abuse? xAI Takes Minnesota’s Nudification Ban to Court

Minnesota’s nudification law was scheduled to take effect on Saturday, August 1. Days before that deadline, Elon Musk’s artificial intelligence company filed a federal lawsuit asking a judge to stop it.

The case matters well beyond Minnesota. It is shaping up as an early test of a question courts have not yet answered clearly: how much authority does a state have to regulate the tools that generate AI imagery, as opposed to the people who misuse them?

What the Law Does

The statute defines nudification narrowly in one sense and broadly in another. It covers images or videos altered or generated to depict an intimate part of an identifiable person that was not present in the original.

The critical design choice is who bears the liability. Most existing laws in this area — state deepfake statutes and the federal Take It Down Act passed last year — target the individuals who create or spread non-consensual sexual imagery, and require platforms to maintain takedown procedures. Minnesota went a step further and aimed at the operators of the services themselves.

Penalties run to $500,000 per violation.

The bill was not a partisan squeaker. It cleared the state House 132 to 1 and the Senate unanimously, 65 to nothing, in a legislature otherwise closely divided. It was signed in May.

What xAI Is Arguing

The 38-page complaint, filed Monday in federal court against Attorney General Keith Ellison, opens with a concession. The company says it does not dispute the state’s interest in prohibiting distribution of AI-generated nude images of real people without their consent.

Its objection is to scope. The filing characterizes the statute as an overbroad, content-based restriction on expression and on the tools of visual creation — a clumsy approach, in its words, to a legitimate problem.

The specific complaints:

• No safe harbor exists for companies making good-faith efforts to block misuse
• The prohibition reaches images the depicted person consented to, and even images that person created themselves
• The definition of an intimate part is written broadly enough to capture body parts routinely visible in public
• The per-violation penalty structure creates exposure the company estimates could theoretically reach $50 billion

That last figure assumes roughly 100,000 violations, a number that arguably illustrates the scale of the underlying problem as much as it illustrates the severity of the penalty.

The company also describes its own safeguards. Grok’s terms of service bar illegal, harmful or abusive uses that violate someone’s privacy, explicitly including altering a person’s image to depict them nude or sexually. xAI says it enforces this through account suspensions and terminations, reports suspected child sexual abuse material, and has itself sued users who circumvented its technical blocks.

Following a global backlash over sexualized images of women and children generated through its systems, the company said in January it had implemented geo-blocking, preventing image editing of real people into revealing clothing in jurisdictions where that is unlawful. An earlier permissive setting had drawn objections from governments around the world.

Minnesota’s Response

Ellison said his office had not yet been served or read the filing, but did not hesitate on the substance. He described using AI to generate nude images of people against their will as appalling, said there are many worthwhile debates to be had about AI policy and that this is not one of them, and argued the practice strips targets of their dignity while inflicting serious emotional, personal and professional harm.

State Senator Erin Maye Quade, who authored the measure, said the law protects First Amendment rights rather than infringing them, and criticized the makers of the technology for prioritizing profits over protection from image-based sexual abuse.

Governor Tim Walz was blunter still, posting a four-word message on social media inviting the company to court.

Legislators have pointed to a specific driver behind the near-unanimous vote: nudification tools have become a feature of school cyberbullying, with teenage girls among the most frequently targeted.

The Doctrinal Question

Both sides have real arguments, and the outcome is genuinely uncertain.

Content-based restrictions on speech face strict scrutiny in American courts, which is a high bar. Overbreadth doctrine also allows a challenger to attack a law for the protected expression it sweeps in, even if the core conduct being targeted could be constitutionally prohibited. Those are the doors xAI is walking through.

Against that, states have long regulated the production and distribution of certain categories of imagery, and the harm here is concrete rather than speculative. The legislature’s near-unanimity may also matter to a court weighing whether the law reflects a considered judgment about a real problem.

The track record so far is mixed. California’s election deepfake law was blocked on First Amendment grounds, and xAI’s earlier challenge to it was partly successful. Minnesota’s own law survived a separate challenge from Musk’s company in 2025.

An Awkward Backdrop

The lawsuit arrives while xAI faces a proposed class action alleging its own platform was used to generate non-consensual sexualized imagery. The Minnesota complaint also asserts the company shared information with authorities about users who abused the tool — a claim contested elsewhere.

That juxtaposition invites the obvious question critics have raised: why fight a law targeting conduct the company says it already forbids?

The company’s answer would presumably be that the objection is not to the prohibition but to being held strictly liable, without a safe harbor, for what determined users do despite its controls — a position with a long lineage in internet law, whatever one thinks of its application here.

What Comes Next

A judge will likely have to decide quickly whether to pause enforcement while the case proceeds. Texas has taken a different route, attaching liability to operators who knew consent was absent or failed to remove flagged material promptly — a knowledge-based standard that may prove more durable if Minnesota’s flat prohibition falters.

Whichever way this goes, other legislatures are watching. The ruling will tell them how much room they have.