The Appeal Court Blinked First: Eighth Circuit Pauses Minnesota’s AI Nudification Ban as xAI’s First Amendment Fight Escalates

On Friday, October 2, 2026, the Eighth U.S. Circuit Court of Appeals handed Elon Musk’s xAI a victory that a lower court had twice denied: an injunction temporarily blocking Minnesota from enforcing HF-1606, the nation’s first law banning AI “nudification” technology, while the company’s constitutional challenge proceeds. The order does not decide whether the law is constitutional. It does something almost as consequential — it freezes, for the duration of the litigation, the most aggressive state attempt yet to regulate an AI image capability itself rather than the people who misuse it.

For the two months the law had been in force, Minnesota had been the live experiment in whether a state could hold AI companies liable for what users do with their tools. That experiment is now paused, and the template that roughly two dozen states were watching has been sent back to the bench.

What the Eighth Circuit actually did

The ruling is an emergency-stage injunction, not a merits decision. xAI had asked the appeals court in September to enjoin the law pending appeal, after U.S. District Judge Donovan Frank rejected the company’s requests for emergency relief twice — first at the end of July, before the statute took effect August 1, and again in a September order denying a preliminary injunction. Judge Frank’s reasoning was procedural but pointed: xAI had not shown it would suffer irreparable harm while the case continued, and it had waited too long — roughly three months — to seek an emergency order.

The Eighth Circuit’s October 2 order changes the result of that immediate fight. Minnesota had urged the appellate panel to leave the law in place while the appeal ran. The court granted xAI’s motion instead, pausing enforcement against the company for the duration of the litigation. Neither side immediately commented when Reuters reached out after the order landed.

Three things about the order deserve emphasis. First, it is temporary by design: the pause holds while xAI’s lawsuit alleging the law is unconstitutional continues in the lower court and on appeal. Second, it signals that a majority of the appellate panel found xAI’s arguments at least serious enough that enforcing the law mid-litigation risked harm the court could not undo — a meaningful shift from Judge Frank’s finding. Third, and most practically, it means the clock on Minnesota’s civil-penalty regime — up to $500,000 per violation — stops running against xAI while the constitutional questions are resolved.

The law at the center: a ban on the tool, not the user

To understand why this fight has drawn national attention, it helps to be precise about what HF-1606 does. Most state deepfake statutes regulate outputs or intent — prohibiting, say, nonconsensual sexual deepfakes used for harassment, or deceptive election media. Minnesota’s law, which took effect August 1, targets the capability itself: it prohibits website operators, software developers, and platform owners from allowing users to create realistic images showing intimate body parts that did not appear in the original photograph of an identifiable person. Liability attaches to the service that enables the transformation, not only to the person typing the prompt.

The statute was passed by overwhelming bipartisan margins — 132-1 in the House and 65-0 in the Senate — after hearings that included testimony about a man who used ordinary social media photos to generate sexualized images of more than 80 women he knew personally. State Senator Maye Quade, who led the push, framed the law as national leadership: “Today, we led the nation protecting women, children and everyone in public life from the harm caused by AI nudification technology.”

Minnesota Attorney General Keith Ellison, defending the law in court, was blunter about the stakes. In an August filing he wrote that xAI’s Grok Imagine had created “an unparalleled marketplace for digital sexual violence that poses virtually no barrier to entry,” and argued that if the state cannot direct its laws at the technology enabling digital sexual victimization, it is doomed from the start. The state’s core defense rests on the law’s purpose: combating AI-generated child sexual abuse material, an area where the record includes a January watchdog estimate that Grok generated more than 23,000 sexualized images of children over just 11 days, and where three California minors joined a federal class action in March alleging their photos were turned into AI-generated CSAM using Grok.

xAI’s counters are constitutional and practical. The company argues HF-1606 is a content-based restriction on protected speech that will sweep in artistic and expressive uses, and that the statute lacks any safe harbor: liability attaches even if the depicted person consented, even if they created the image themselves, and even if the image was never shared. In its appellate papers, xAI also told the court that Grok Imagine now includes “rigorous protections” against creating sexualized images of real people — noting it has suspended more than 50,000 accounts and filed more than 70,000 reports to the National Center for Missing and Exploited Children in 2026 alone. That is the company’s own characterization of its controls, not an independent assessment, and Minnesota has seized on Grok’s documented history to dispute it.

Why an appellate pause matters more than it sounds

Emergency injunctions are routine procedural events in most litigation. In a case that is functioning as a national test, they are scoreboard. Three dynamics make this one worth watching closely.

The template question. Roughly two dozen states have enacted some form of deepfake or AI imagery law, but Minnesota’s was the first to target the nudification capability itself. If the approach survives First Amendment scrutiny, copycat legislation is widely expected. If it falls, states will be pushed toward narrower statutes that regulate distribution and intent rather than the tool. The Eighth Circuit’s willingness to pause enforcement is an early indicator — not a holding — of how that fight may lean.

The tool-versus-speech framing. The deepest legal question in the case is whether nudification software is “speech” protected by the First Amendment or a “tool” the state may regulate like any other dangerous product. Precedent cuts both ways: courts have blocked state deepfake laws, including California’s, when they reached political speech or satire, but regulators have historically prevailed when a law targets a tool’s capacity for harm rather than a viewpoint. The panel’s decision to pause a statute of the second kind suggests at least some appetite on the bench for taking the speech theory seriously.

The enforcement gap. Every month a tool-targeting law is paused is a month the conduct it prohibits — including, unambiguously, CSAM generation using photos of real children — continues where a company declines to prevent it. Minnesota built its defense around exactly that gap. The pause does not invalidate the state’s argument, but it extends the period during which the harms the legislature was responding to remain legally unaddressed at the state level, with enforcement left to federal CSAM law and platform self-policing.

What happens next

The litigation now proceeds on two tracks. In the district court, Judge Frank continues to shepherd the case toward a merits ruling on the constitutional challenge — the order xAI ultimately needs. In the Eighth Circuit, the September appeal of the denial itself will be briefed and argued, and the October 2 pause makes it materially less likely that the law operates against xAI at any point before the appellate decision lands.

For AI developers, the practical takeaway is narrower than either side’s rhetoric. The case has not established that AI image tools are speech, nor that states cannot regulate them. It has established that a well-resourced company willing to litigate aggressively can freeze a first-of-its-kind state AI statute within two months of its effective date — and that the First Amendment is a live weapon in that fight. Companies building image-generation products with U.S. exposure will read the order as breathing room; state legislators drafting the next wave of AI imagery bills will read it as a warning that tool-targeting bans face a real risk of years in litigation before they ever bind anyone.

The losers in the interim are the people the Minnesota Legislature had in mind when it voted 132-1. However the First Amendment analysis resolves, the pause guarantees a longer window in which the harms documented in that legislative record — and the ones in Ellison’s filings — sit outside that state’s reach. The court did not say those harms don’t matter. It said the constitutional questions are close enough, and the company’s position credible enough, that the law must wait. Whether that patience is vindicated or regretted will be decided in the briefing to come — and, eventually, by whatever precedent this case becomes for every state now writing its own version of HF-1606.