Minnesota AI ban survives challenge

A live test for state AI enforcement

Minnesota’s ban on AI “nudification” survived its latest court challenge on September 4, when U.S. District Judge Donovan W. Frank denied xAI’s request for a preliminary injunction and allowed the law to remain in force while the company’s lawsuit proceeds [1]. The decision does not end the case, and it does not finally decide whether the statute is constitutional. But it does something immediately important: it keeps Minnesota’s enforcement model alive at the very moment courts, platforms and lawmakers are trying to define who is responsible when generative-AI tools are used to create realistic sexual images of real people.

The law, known during the legislative process as H.F. 1606 and codified as Minnesota Statutes section 325E.91, bars owners or controllers of websites, applications, software, programs or other services from allowing users to “nudify” an image or video, or from doing so on a user’s behalf [1]. The order describes “nudify” as altering or generating an image or video so that it realistically depicts an intimate part of an identifiable person that was not present in the original image or video [1]. For violators, the statute allows the Minnesota attorney general to seek civil penalties of up to $500,000 per violation, along with injunctive relief and other remedies [1].

That exposure is why the case matters beyond Minnesota. The state is not simply criminalizing downstream revenge-porn distribution after an image has already spread. It is targeting the providers of tools that can enable the creation of those images in the first place. In other words, Minnesota is testing whether AI content-safety obligations can be imposed by state law, not just promised in platform policy pages.

Why the judge refused to freeze the law

Judge Frank’s ruling turned heavily on timing and harm. H.F. 1606 was signed into law on May 7, 2026, and took effect on August 1, 2026; xAI filed its lawsuit and emergency request near the end of July, just days before enforcement was set to begin [1]. The court found that the company’s delay undermined its claim that it faced immediate, irreparable harm if the statute stayed in effect [1].

The judge also rejected xAI’s argument that possible penalties or compliance costs justified a preliminary injunction at this stage. The order noted that any $500,000 penalty would be monetary in nature and could be recovered if the law were ultimately invalidated, and it said xAI had already implemented Minnesota-specific controls in Grok Imagine’s image-editing features [1]. According to the court, xAI’s evidence of future compliance costs and lost users was too vague and speculative to satisfy the demanding standard for emergency relief [1].

That does not mean the judge brushed aside the First Amendment issues. Frank wrote that the constitutional questions are complex and will receive fuller consideration as the case moves forward, including through Minnesota’s pending motion to dismiss [1]. But for now, he concluded that xAI had not shown the kind of irreparable harm required to stop a democratically enacted statute before the merits are decided [1].

The public-interest analysis was just as significant. The court said the balance of harms and public interest “tips steeply” toward Minnesota, emphasizing the state’s interest in responding to the undisputed harm caused by AI-nudification technology and the proliferation of photorealistic sexualized images [1]. MPR News reported that the decision permits Minnesota regulators to impose the statute’s fines while the case continues, and that xAI had previously told the court it would suspend affected features for Minnesota users because of enforcement concerns [2].

The evidence of harm behind the statute

The ruling gives unusual weight to the real-world record developed before Minnesota lawmakers. According to the order, three women testified at a February legislative hearing about extensive harms after a perpetrator used nudification technology to create realistic pornographic material using their likenesses and the likenesses of more than 80 women [1]. The court summarized harms including missed work, fear for family members and children, fear of leaving home, and the continuing distress of not knowing who had seen or might later see the images [1].

The order also cited material presented to legislators showing the scale of the problem. RAINN told lawmakers that, in one month in 2023, nudification apps had 24 million users, and that more than 95 percent of manipulated images in 2023 were nonconsensual, sexually explicit and featured women [1]. The court also pointed to evidence involving children, including reports from the Internet Watch Foundation about AI-generated child sexual abuse material and the use of nudification tools for blackmail [1].

A separate legislative record cited in the order found that, during the first 11 days after a social-media AI editing tool launched, about 65 percent of generated photos were photorealistic sexualized images, totaling roughly 3 million images at a pace of about 190 per minute; around 23,000 of those sexualized images were of children [1]. Valley News Live reported the same figure from the ruling and noted that the court framed those facts as part of the harm Minnesota sought to prevent [3].

xAI’s free-speech argument remains alive

xAI’s broader lawsuit is still pending. The company argues that the Minnesota law violates the First Amendment, both facially and as applied, and that the statute restricts tools of visual expression too broadly [1]. In the court record, xAI also says Grok Imagine can be used to create many kinds of lawful images, including art, family images, memes, product mockups, political satire and religious imagery [1].

The company maintains that its terms of service prohibit illegal, harmful or abusive conduct, including nudification of a real person’s image, altering a real person’s image to depict them in an intimate or sexual context, and sexualizing or exploiting children [1]. The order says xAI described enforcement measures such as suspensions, terminations and reports of suspected child sexual abuse material to law enforcement [1].

Minnesota’s position is that those private policies are not enough. The state argued, and the court credited at this stage, that the harms of AI nudification are distinct because creation, replication and circulation can happen quickly and because takedown remedies after distribution may be inadequate [1]. Courthouse News reported that Frank viewed the statute as a direct response to harms unique to the technology, rather than merely an attempt to punish individual users after images are already circulating [1].

Reuters reported that xAI failed on September 4 to persuade the federal judge to block the law, which applies to website operators, software developers and others that allow users to nudify images of identifiable people using AI technology [4]. That framing is central: Minnesota has aimed the statute at providers that control access to the tool, not only at the person typing the prompt.

Why this ruling matters for AI companies

For generative-AI companies, the practical message is direct. The era in which image-generation safety could be treated mainly as an internal trust-and-safety choice is narrowing. If Minnesota’s approach survives later stages of litigation, companies offering image or video tools may need jurisdiction-specific safeguards, clearer audit trails, stronger refusal systems and faster detection of attempts to generate intimate images of identifiable people.

The ruling also highlights a growing divide between voluntary platform governance and legally enforceable safety duties. xAI told the court it had implemented additional controls in Minnesota as of August 1 and that moderation rates for image-editing prompts in the state had risen by more than 10 percent [1]. The company presented that as evidence of burden; the court treated it as insufficient proof of irreparable harm, especially because the adjustments had already been made [1].

That distinction could shape future AI litigation. Courts may be sympathetic to constitutional challenges that receive full merits review, but they may be less willing to freeze state safety laws when companies delay, when asserted compliance harms are generalized, or when the public record shows concrete harms to victims. The Minnesota case therefore becomes an early blueprint for other states: define the prohibited synthetic conduct, target the provider’s ability to enable it, document harms in the legislative record, and defend the law as a consumer-protection or civil-enforcement measure.

Still, the legal fight is far from over. Judge Frank expressly left the First Amendment merits for later consideration, and the state’s motion to dismiss will put those issues “front and center,” according to the order [1]. The next phase will test whether Minnesota’s statute is narrow enough to survive constitutional scrutiny while still strong enough to meaningfully curb AI-generated sexual abuse.

For now, however, the result is clear. Minnesota’s AI nudification ban remains enforceable. Musk’s AI company did not obtain the pause it sought. And one state now has a live legal model for moving AI image abuse from the realm of platform promises into the realm of enforceable law.

Sources from the last 72 hours

1. [1]Memorandum Opinion and Order, X.AI LLC v. Keith Ellison, Civil No. 26-3425Sep 4, 2026, 5:00 AM UTC
2. [2]Federal court denies X.AI’s bid to block Minnesota’s AI nudification banSep 4, 2026, 7:34 PM UTC
3. [3]Musk’s xAI loses court bid to block Minnesota's AI ‘nudification’ banSep 4, 2026, 4:08 PM UTC
4. [4]Judge lets Minnesota enforce anti-nudification app law over xAI objection as case proceedsSep 4, 2026, 4:27 PM UTC

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.