Minnesota’s new prohibition on automated artificial-intelligence tools that create realistic nude images of identifiable people took effect Saturday, August 1, after a federal judge declined to grant xAI an emergency order suspending the statute.

The decision leaves the law operational while the court considers a broader request from xAI for a preliminary injunction. The company, controlled by Elon Musk and responsible for the Grok chatbot and image-generation system, filed its constitutional challenge only days before the statute’s effective date. It argues that Minnesota has written an excessively broad restriction that places unlawful burdens on protected expression and exposes technology providers to extraordinary financial penalties.

U.S. District Judge Donovan Frank denied xAI’s request for a temporary restraining order on July 31. According to the court schedule reported by Minnesota media, Frank concluded that the company’s delay in bringing the case weakened its assertion that it faced an immediate and irreparable injury requiring emergency intervention. The statute was signed by Gov. Tim Walz on May 7, giving affected companies nearly three months before enforcement began.

The ruling was procedural rather than a final determination that Minnesota’s law is constitutional. Frank directed the parties to address xAI’s motion under the more developed preliminary-injunction process. Minnesota Attorney General Keith Ellison is expected to submit the state’s opposition by August 12, followed by xAI’s final brief on August 17. A hearing is scheduled for August 19.

That distinction is significant for technology companies evaluating the decision. The court has not yet ruled on whether the statute impermissibly regulates speech, whether its definitions are too broad or whether its liability provisions conflict with federal protections for online platforms. It has instead allowed enforcement to begin while those questions remain under review.

The Minnesota statute, enacted as Chapter 72 of the state’s 2026 session laws, targets the providers and promoters of nudification technology rather than focusing exclusively on individuals who create or distribute abusive images. It applies to a person or company that owns or controls a website, application, software program or other service and allows users to access, download or use that service to nudify an image or video.

The law also prohibits a provider from performing the nudification on behalf of a user. Separately, it bars advertising or promoting a website, application, program or service that performs the prohibited activity. That structure gives the measure a substantially wider commercial reach than laws directed only at the publication or distribution of nonconsensual intimate imagery.

Under the statute, “nudify” refers to altering or generating an image or video so that it depicts an intimate body part that did not appear in the original material, when the result is realistic enough that a reasonable person could believe the depicted body part belongs to the identifiable individual. A person may be identifiable from the image itself, by someone who recognizes the person or through personal information displayed with the image.

The legislation includes an exemption when the relevant software requires substantial individualized technological or artistic skill and judgment from a human user. That provision appears intended to distinguish automated consumer tools from professional software requiring meaningful manual control. Its boundaries, however, could become an important issue as generative-image platforms increasingly combine text prompts, automated editing and more advanced user-directed workflows.

Minnesota has established both private and government enforcement mechanisms. An individual depicted in a prohibited nudified image may file a civil action and seek compensatory damages, including damages for mental anguish or suffering, in an amount reaching three times the actual loss. Plaintiffs may also request punitive damages, an injunction, attorney fees, litigation costs and other relief considered appropriate by a court.

The attorney general may independently enforce the law and seek a civil penalty of as much as $500,000 for each unlawful access, download or use of the covered technology. The wording makes the number of potential violations a central financial risk. A platform experiencing repeated prohibited uses could theoretically face liability calculated on a per-use basis rather than through a single penalty covering the service as a whole.

Money collected through state civil penalties is to be directed toward grants supporting organizations that provide services and advocacy for victims of sexual assault, domestic violence, child abuse and other crimes. The statute also establishes Minnesota jurisdiction when either the plaintiff or defendant resides in the state and allows an affected resident to bring an action in the county where that person lives.

xAI’s lawsuit does not dispute that Minnesota has a legitimate interest in preventing the creation and dissemination of nonconsensual synthetic nude images. Its challenge instead centers on the method and breadth of regulation. The company contends that the statute prohibits more expression than is necessary to address the state’s stated objective.

Among other arguments, xAI says the law could cover images created with the depicted adult’s permission, including material created by the depicted individual. It also argues that liability may attach even when an image remains private and is never distributed, distinguishing the Minnesota measure from laws focused on public dissemination or refusal to remove harmful material after notice.

The company further objects to the statutory definition of an intimate part, contending that it may encompass parts of the body that can lawfully appear in public or nonsexual contexts. In xAI’s account, the prohibition therefore risks covering artistic, humorous, political or otherwise protected visual expression in addition to the nonconsensual sexual imagery that lawmakers intended to combat.

Another major point of contention is the absence of a broad safe harbor for providers that deploy moderation systems and attempt in good faith to prevent prohibited outputs. xAI says its services prohibit illegal, harmful and abusive uses, including the alteration of a real person’s image to create nudity or sexually explicit material without permission. The company says it can suspend or terminate users who violate those policies and report suspected child sexual-abuse material.

Those internal rules do not necessarily eliminate exposure under the Minnesota statute. The law focuses on whether an owner or operator allows users to access or use a service for nudification, creating uncertainty over the level of filtering, testing and enforcement that would be sufficient. A provider could argue that a user circumvented sophisticated controls, while the state or a private plaintiff could argue that the service remained technically capable of producing the prohibited result.

For developers of general-purpose image models, the compliance challenge is broader than removing a product explicitly marketed as an “undressing” application. Modern systems can edit uploaded photographs, synthesize realistic bodies, transform clothing and generate new scenes through ordinary-language prompts. Regulators may therefore examine not only how a company describes its product but also what users can accomplish through combinations of prompts, reference images and editing functions.

The immediate operational responses available to AI companies include disabling certain image-editing functions for Minnesota users, blocking uploads of photographs depicting real people, applying stronger prompt and output classifiers, limiting adult-content features or geofencing an entire service. Each approach carries costs and technical limitations. Geographic restrictions can be evaded, automated filters may block lawful material, and a statewide product withdrawal can reduce revenue and user growth while attracting criticism from customers.

Providers may also need more extensive records demonstrating how moderation systems were designed and enforced. Relevant documentation could include safety testing, rejected prompts, detection thresholds, user sanctions, geographic controls, complaint handling and model updates. Although the Minnesota law does not expressly establish a compliance certification regime, such records could become important in litigation over whether a platform knowingly facilitated prohibited conduct or responded adequately to misuse.

The commercial consequences extend beyond model developers. App stores, cloud providers, payment companies, online advertising networks and social platforms may need to evaluate whether their relationships with nudification services create reputational, contractual or regulatory risks. The statute expressly restricts advertising and promotion, increasing the likelihood that marketing intermediaries will review clients offering automated image-manipulation products.

Minnesota’s approach differs from federal and state laws that primarily target people who distribute nonconsensual intimate images or require platforms to remove such content after receiving a valid request. The federal TAKE IT DOWN Act established obligations concerning nonconsensual intimate imagery and removal procedures, but Minnesota is attempting to intervene earlier in the technology chain by restricting access to tools capable of producing the images.

Texas has adopted a different provider-liability structure tied more closely to knowledge, consent and responses to reported material. Minnesota’s law is potentially more restrictive because it broadly prohibits covered automated services and does not make liability depend solely on whether the operator knew that a particular depicted person had withheld consent.

The case consequently raises a core question for emerging AI regulation: whether governments may prohibit a technological capability because of its predictable harmful uses, or whether they must target specific unlawful users, outputs and acts of distribution. xAI maintains that Minnesota has regulated too much protected expression in attempting to prevent abuse. The state argues that readily accessible nudification tools create a distinct category of harm that cannot be adequately addressed only after images have been created or circulated.

Attorney General Ellison has defended the measure as a protection for the dignity and safety of Minnesota residents. Supporters of the law say realistic synthetic sexual images can produce severe emotional, personal and professional harm even when the images are known to be fabricated. The speed and low cost of generative tools can also allow a single user to create many images from ordinary photographs obtained through social-media accounts.

The legislation received overwhelming bipartisan support in Minnesota. State Sen. Erin Maye Quade, one of its leading advocates, has said the proposal grew from accounts involving women whose publicly available photographs were transformed into realistic sexual images. For lawmakers, those cases demonstrated that conventional remedies aimed at distribution or harassment could leave a gap when automated services make the underlying production process widely accessible.

xAI’s challenge arrives amid broader scrutiny of Grok’s image-generation features. The company faced international criticism after users generated sexualized images involving real people, including reports of images involving minors. xAI subsequently said it had implemented technical controls intended to prevent the editing of photographs of real people into revealing clothing and would geographically block content where local law prohibited it.

The Minnesota litigation will test how courts assess those safeguards when a state chooses to regulate the provider itself. If xAI ultimately secures an injunction, the decision could discourage similarly broad product bans and push lawmakers toward narrower rules based on nonconsent, knowledge, distribution or failure to remove content. A ruling for Minnesota could encourage other states to impose direct obligations on AI model operators and application developers.

The constitutional analysis is likely to examine whether the law is a content-based restriction on expression, whether its definitions provide adequate notice, whether less restrictive alternatives are available and whether the statute burdens a substantial amount of lawful activity. Questions involving federal preemption and Section 230 protections for online intermediaries may also emerge, although the Minnesota law states that it should be construed consistently with existing federal liability protections.

Even without a final judgment, the August 1 effective date changes the risk calculation for companies operating in Minnesota. Platforms can no longer assume that a terms-of-service prohibition alone will satisfy state authorities or prevent private claims. Companies must decide whether their technical controls make prohibited outputs sufficiently inaccessible, whether additional restrictions are necessary and whether continued operation of certain features is commercially justified.

The next major development is expected at the August 19 preliminary-injunction hearing. Unlike the temporary restraining-order decision, that proceeding should give the court a fuller record on the constitutional merits, the scope of potential harm to xAI, the state’s enforcement interests and the practical effect of leaving the statute in force during litigation.

Until the court rules again, Minnesota’s prohibition remains active. The result places xAI and other covered technology providers under a novel state liability regime while establishing Minnesota as an early testing ground for direct regulation of generative-AI capabilities. The eventual decision could influence how lawmakers across the United States draw the line between regulating harmful digital conduct and restricting the software systems that make such conduct possible.