Court Keeps Minnesota AI Nudification Law in Force

A federal judge has denied xAI’s request for a preliminary injunction against Minnesota’s AI nudification law. The September 4, 2026 order leaves the statute enforceable while xAI’s constitutional lawsuit continues, but it does not finally decide whether the law violates the First Amendment.

What the federal court decided

Judge Donovan W. Frank of the U.S. District Court for the District of Minnesota refused to pause enforcement of the state law for the duration of xAI’s lawsuit. The case is X.AI LLC v. Ellison, No. 0:26-cv-03425. The September order addresses provisional relief at an early stage, not xAI’s ultimate request to have the law declared unconstitutional and permanently blocked.

A preliminary injunction is an extraordinary remedy issued before a case reaches final judgment. The applicant generally must show, among other factors, a sufficient threat of irreparable injury, a likelihood of success and that the balance of harms and public interest support intervention. Reporting on the order shows the court concluded that xAI did not establish the required immediate, irreparable harm and that the balance of harms and public interest strongly favoured Minnesota.

The court also returned to xAI’s timing. Minnesota Governor Tim Walz signed the legislation on May 7, but xAI sued on July 27 and sought emergency relief on July 29, three days before the August 1 effective date. Judge Frank wrote that a sophisticated, well-resourced litigant genuinely fearing irreparable injury could have acted sooner.

That finding matters procedurally. Courts distinguish injury that can be addressed later—such as some monetary loss—from harm that cannot be repaired after final judgment. Bloomberg Law reported that the court viewed potential penalties as recoverable if xAI ultimately succeeds and found the company’s evidence about moderation costs and lost users vague or insufficiently concrete.

xAI v. Ellison: verified procedural facts

Event	Date	Legal significance
Governor signs H.F. 1606	May 7, 2026	Creates Minnesota Statutes section 325E.91
xAI files lawsuit	July 27, 2026	Begins constitutional challenge in federal district court
xAI seeks emergency relief	July 29, 2026	Requests a TRO and preliminary injunction
TRO denied	July 31, 2026	Law is allowed to take effect the next day
Law becomes effective	August 1, 2026	Covered conduct after this date can trigger civil enforcement
Preliminary-injunction hearing	August 19, 2026	Court hears the broader request to pause enforcement during litigation
Preliminary injunction denied	September 4, 2026	Law remains enforceable; merits case continues

Why the September ruling is different from the July TRO

The July and September orders are related but not interchangeable. A temporary restraining order, or TRO, is short emergency relief designed to preserve the situation until a court can consider a fuller injunction request. On July 31, the court denied xAI’s TRO bid, pointing to the company’s delay and lack of a need for immediate judicial action. Minnesota’s law therefore took effect on August 1.

The court then considered a preliminary injunction after briefing and an August 19 hearing. That request sought to stop enforcement for a longer period while the lawsuit proceeds. The September 4 denial means Minnesota can continue enforcing the statute unless an appellate court or a later district-court ruling changes that position.

Neither order is the same as a final judgment after full merits consideration. The September decision did not hold that every constitutional objection is invalid. Reuters and MPR reported that Judge Frank called the issues complex and said they deserved full consideration. MLex likewise reported that the judge did not finally decide the First Amendment question.

This distinction also affects the word “wins.” Minnesota prevailed on the interim motion, but the state has not received a final judgment upholding the statute against all of xAI’s claims. xAI lost its present attempt to pause enforcement; it has not exhausted its challenge.

What Minnesota’s law actually prohibits

The enacted text is Minnesota Laws 2026, chapter 72, codified at section 325E.91. It applies to a person who owns or controls a website, application, software, programme or other service. That operator must not allow a user to access, download or use the service to “nudify” an image or video, and must not perform that process for a user. The law also prohibits advertising or promoting a service that performs the covered actions.

The statute defines “nudify” more narrowly than any digital alteration. The output must depict an “intimate part” that was not shown in the original image or video of an identifiable individual, and it must be realistic enough that a reasonable person would believe the depicted body part belongs to that person. The definition of an identifiable individual can be satisfied by the image itself, recognition by the depicted person or another person, or personal information displayed with the image.

An exemption applies when the service requires “technical skill,” defined as substantial application of individualised technological or artistic skill and human judgment in directing, shaping or controlling the output. The scope and application of that exemption are likely to matter as the case continues.

The law provides two civil-enforcement routes. A depicted individual may sue for compensatory damages—including mental anguish—up to three times actual damages, punitive damages, injunctive relief, legal fees and other appropriate relief. Separately, the Minnesota attorney general may enforce the statute and seek a civil penalty of up to $500,000 for each unlawful access, download or use covered by subdivision 2.

The $500,000 figure is a maximum civil penalty, not an automatic fine already imposed on xAI. No source reviewed for this article reports that Minnesota has assessed such a penalty against the company. The statute also says it does not alter federal liability protections and must be construed consistently with Section 230 of the federal Communications Decency Act.

What xAI argues

xAI’s complaint contends that Minnesota adopted an overbroad, content-based restriction on speech and expressive tools. The company argues the statute reaches beyond non-consensual sexual imagery and can cover consensual, non-disseminated or non-sexual expression. It also objects to the absence of a broad safe harbour for a provider whose rules and technical controls prohibit misuse but whose users find a way around them.

The company has said its policies bar illegal, harmful or abusive uses, including non-consensual nudification and child sexual exploitation. Its legal position is not that the state lacks any legitimate interest in preventing non-consensual intimate imagery. The dispute is whether the mechanism and breadth of section 325E.91 comply with the First Amendment and other federal law.

Minnesota responds that the statute regulates access to products that automate a uniquely harmful process and that the state can protect residents from realistic, non-consensual sexualised images. The attorney general also argues xAI cannot transform a claimed possibility of future compliance costs into the immediate irreparable injury required for preliminary relief.

The court’s September order leaves those positions for further litigation. It is therefore inaccurate to say the ruling established that AI output is or is not protected speech in all circumstances. The result is narrower: xAI did not meet the requirements for a preliminary injunction on the record before the court.

Why the court focused on irreparable harm

Preliminary relief is not granted simply because a plaintiff raises a serious constitutional argument. The applicant must also show why waiting for ordinary litigation would cause injury that a later ruling cannot repair. According to Reuters and Bloomberg Law, xAI pointed to exposure to penalties, compliance and moderation work, and possible loss of users or features in Minnesota.

The court was not persuaded that those claimed injuries were sufficiently immediate or irreparable. It noted that financial harms can sometimes be remedied later and that xAI had already made changes to its tools without presenting concrete evidence of future cost. The timing of the lawsuit further weakened the claim of urgency.

On the other side, the court credited Minnesota’s public-interest case. The legislature passed the measure 132–1 in the House and 65–0 in the Senate after hearing about harms to people depicted in synthetic sexual imagery. The order treated the prevention of those harms as weighty while litigation continues.

This balancing does not immunise the statute from later constitutional review. A court can allow a law to remain effective at the preliminary stage and later invalidate part or all of it after a fuller record. Conversely, the denial could survive appeal and the state could ultimately prevail. Both paths remain legally possible.

What happens next

MPR and Reuters reported that xAI filed notice it would seek review by the U.S. Court of Appeals for the Eighth Circuit. An appeal could challenge the denial of preliminary relief without immediately resolving the entire underlying case. The appellate court could affirm, reverse, modify the ruling or consider a request for a stay while it reviews the dispute.

Back in the district court, Minnesota has a motion to dismiss pending. If the case survives dismissal, the parties may develop a fuller record about how the statute operates, the technical safeguards available to generative-AI providers, the types of expression covered and the alleged burdens on protected speech.

The next useful documents are therefore appellate filings, any stay request, the Eighth Circuit’s response, the district court’s decision on dismissal and eventually a merits order. Claims that the law has been permanently upheld—or permanently invalidated—should wait for those records.

Why AI companies should pay attention

Minnesota’s approach targets the provider of an accessible tool, not only a person who creates or distributes a harmful image. That design creates operational questions for general-purpose image generators: providers need to determine which features fall within the definition, how location controls work, how repeated misuse is detected and what evidence can demonstrate reasonable safeguards.

The statute’s “technical skill” exemption also draws a line between automated consumer tools and software requiring substantial human direction. Product teams should not assume that an ordinary prompt constitutes technical skill under Minnesota law. The enacted definition refers to substantial, individualised technological or artistic skill and human judgment.

For Indian founders and global AI operators, the case is a reminder that product compliance increasingly changes by jurisdiction. Lapaas Voice’s overview of AI regulation and ethics in India shows how policy models vary across markets. Our report on India’s developing deepfake rules addresses a different legal system but the same need to distinguish creation, distribution, labelling and platform responsibility.

The Minnesota dispute also sits beside wider litigation involving model providers. The xAI Grok training-data lawsuit concerns different claims and should not be conflated with this constitutional challenge. Likewise, our report on the EU AI Act’s deepfake labelling rules covers transparency duties rather than Minnesota’s product-access prohibition.

Frequently asked questions

Did the court uphold Minnesota’s law as constitutional?

No final constitutional judgment has been entered. The court denied xAI’s request for a preliminary injunction, leaving the statute enforceable while the constitutional case continues.

What happened in July 2026?

Judge Frank denied xAI’s emergency temporary restraining order on July 31. That narrower ruling allowed the law to take effect August 1 while the court considered the longer preliminary-injunction request.

What did the court decide on September 4?

The court denied the preliminary injunction. It found xAI had not demonstrated the required irreparable harm and that the balance of harms and public interest favoured the state at this stage.

Can Minnesota impose a $500,000 fine automatically?

No. The law authorises a civil penalty of up to $500,000 for each unlawful access, download or use, enforceable by the attorney general. The maximum is not an automatic fine and no reviewed source reports that such a penalty has already been imposed on xAI.

Is xAI appealing?

Reuters and MPR reported that the company filed notice that it would seek Eighth Circuit review. Further appellate filings and any request for a stay should be checked on the docket.

Source and verification note

The law’s text, definitions, remedies and effective date were checked against Minnesota Laws 2026, chapter 72. The case is X.AI LLC v. Ellison, No. 0:26-cv-03425 (D. Minn.); the docket is available through CourtListener. The July TRO history and Minnesota’s opposition are documented by the Minnesota Attorney General’s July statement and August filing notice.

Independent reporting on the September order was checked through Reuters, MPR News, CBS Minnesota, FOX 9, MLex and Bloomberg Law. Those sources consistently distinguish the September preliminary-injunction denial from the unresolved merits.

The bottom line

Minnesota may continue enforcing its AI nudification law because xAI failed to obtain preliminary relief. The practical result is important for platforms operating in the state, but the legal conclusion is limited: the judge did not issue a final ruling that resolves xAI’s First Amendment challenge. The next decisive evidence will come from the Eighth Circuit and later district-court merits proceedings.