xAI has filed a federal lawsuit challenging Minnesota’s first-in-the-nation law targeting artificial intelligence tools that can generate fake nude images of identifiable people.
The company behind Grok argues that the legislation goes further than preventing non-consensual sexual imagery. According to xAI, the law also restricts lawful visual expression, covers consensually created images and exposes AI providers to enormous penalties even when they have introduced safeguards against misuse.
Minnesota officials maintain that stronger intervention is necessary because AI “nudification” tools can turn ordinary photographs into realistic sexual images without the knowledge or consent of the person depicted. The legal dispute now creates an important test of how far states can go when regulating generative AI systems that can be used for both legitimate and harmful purposes.
What Minnesota’s New Nudification Law Prohibits
Minnesota’s House File 1606 is scheduled to take effect on August 1, 2026. The law says that a person controlling a website, application, program, software product or other service must not allow users to access, download or use that service to “nudify” an image or video. Providers are also prohibited from performing the transformation on a user’s behalf.
The legislation defines nudification as using an automated process to alter or generate an image or video of an identifiable person so that it depicts an intimate body part that was not visible in the original material. The law focuses on technology requiring little or no technical skill, distinguishing automated AI services from conventional professional editing tools.
The complete wording can be reviewed through the Minnesota Legislature’s official HF 1606 record, while the Minnesota House has published a more accessible summary of the new nudification law.
The legislation also restricts advertising and promotion of prohibited nudification services. Minnesota lawmakers designed it to address the tools enabling the creation of synthetic intimate imagery rather than waiting until the resulting material has already been distributed online.
Violations Could Bring Penalties of Up to $500,000
The financial consequences are one of the most controversial parts of the law. Minnesota’s attorney general may pursue civil penalties of up to $500,000 for each violation.
A person depicted in a generated image may also bring a private civil claim. Available remedies can include compensation for financial loss and emotional distress, punitive damages, injunctive relief and legal fees. Depending on the circumstances, damages related to mental anguish may reach three times the person’s actual damages.
Supporters believe substantial penalties are necessary because synthetic intimate images can be created rapidly, distributed widely and copied repeatedly. Smaller fines might be treated as an acceptable cost of operating a profitable generative AI service.
For xAI, however, the size of the potential penalty creates pressure to block a much broader category of image-generation activity. Further coverage of the dispute is available through the Associated Press report on xAI’s Minnesota lawsuit.
Why xAI Says the Law Violates Free-Speech Rights
xAI’s 38-page complaint argues that HF 1606 creates an overbroad, content-based restriction on speech and visual expression.
The company says it does not oppose laws targeting the non-consensual distribution of fake nude images. Its objection is that Minnesota’s law regulates the creation and availability of the underlying technology regardless of whether the depicted person consented or whether the resulting image was ever publicly shared.
According to the complaint, liability could potentially apply when an adult voluntarily edits an image of themselves, when a person grants permission for an image to be created or when a generated work has artistic, satirical, educational, medical or political significance. xAI argues that these scenarios may involve constitutionally protected expression rather than abuse.
The company also claims that the definition of an “intimate part” is broad enough to include areas of the body commonly visible in swimsuits, shorts or shirtless photographs. It warns that providers could respond by restricting legitimate image-editing functions for Minnesota users.
A detailed explanation of the company’s free-speech argument appears in The Verge’s analysis of the xAI complaint.
The Lack of a Safe Harbor Is Central to xAI’s Challenge
Another major concern is the law’s apparent lack of a safe harbor for technology companies that make good-faith attempts to prevent abuse.
xAI says providers may face liability even after implementing moderation policies, automated filters and other technical controls. A determined user might bypass those protections, yet the company operating the model could still be exposed to substantial penalties.
The lawsuit characterizes the law as imposing strict liability based on what a user manages to produce rather than on whether the company knowingly supported or encouraged the conduct. xAI says such a structure may force platforms to disable useful image-generation capabilities because no moderation system can guarantee that every prohibited output will be blocked.
The company states that Grok’s policies prohibit users from generating nude or sexualized images of people without consent. Whether those protections are sufficient, and whether a provider’s stated policies should reduce its legal responsibility, will likely become important questions as the case progresses.
Minnesota Says the Harm Requires Stronger Protection
Minnesota Attorney General Keith Ellison has defended the purpose of the legislation, describing involuntary AI nudification as a serious violation of personal dignity capable of causing emotional, personal and professional harm.
State Representative Jessica Hanson, who sponsored the House legislation, has argued that victims need meaningful ways to hold both users and technology providers accountable. The proposal received overwhelming legislative support, passing the Minnesota House by 132 votes to one and the state Senate by 65 votes to zero.
Victims who testified during the legislative process described discovering fabricated sexual images created from photographs that had originally been shared for ordinary personal or professional purposes. Lawmakers concluded that rules focused only on distribution may intervene too late because the harm can begin as soon as an intimate image is generated.
Minnesota’s legislative background and testimony can be found through the state House’s official session coverage.
How the Law Differs From the Federal TAKE IT DOWN Act
The federal TAKE IT DOWN Act already addresses non-consensual intimate imagery, including computer-generated deepfakes. It criminalizes certain intentional publications of intimate visual depictions without consent and requires covered platforms to establish a process for removing qualifying content after receiving a valid request.
Minnesota’s law moves further upstream. Rather than concentrating mainly on publication and removal, it targets access to automated services capable of producing the images.
xAI argues that the federal approach is more narrowly connected to non-consensual distribution and includes clearer exceptions. Minnesota supporters respond that a removal-focused system cannot completely repair the damage after an image has already been created, downloaded or circulated.
The exact federal provisions are available in the official text of the TAKE IT DOWN Act. The difference between regulating harmful distribution and regulating the technology used to create harmful material could become one of the most significant issues in the Minnesota case.
The Case Could Influence AI Laws Beyond Minnesota
The lawsuit is larger than a dispute involving one company and one state. It could help determine whether governments may restrict access to a category of generative AI tools based on their capacity for abuse, or whether laws must focus more narrowly on individual users and clearly non-consensual outputs.
A victory for Minnesota could encourage other states to adopt direct restrictions on nudification services. A victory for xAI could push lawmakers toward narrower legislation containing consent exceptions, protected-use exemptions and safe harbors for platforms with effective safeguards.
The dispute also illustrates the difficulty of regulating dual-use AI. The same model may produce harmless edits, satire, artwork and educational illustrations while also being manipulated to create abusive sexual imagery. A law written too narrowly may leave victims unprotected, while one written too broadly may suppress lawful expression and useful technology.
Research has found that many consumer-facing face-swap applications lack meaningful technical protections against sexual misuse, reinforcing arguments that voluntary moderation alone may not adequately address the problem.
A Defining Test for Generative AI Accountability
Minnesota is attempting to stop synthetic sexual abuse at the point of creation. xAI is arguing that the state has chosen a constitutionally excessive method that may punish protected expression and hold providers responsible for conduct they actively prohibit.
Both sides acknowledge the serious harm caused by non-consensual intimate deepfakes. The conflict concerns how that harm should be prevented, which parties should bear responsibility and how much lawful functionality a government may restrict in the process.
The court’s response could shape the next generation of state AI laws. It may determine whether broad technology bans can survive constitutional review or whether lawmakers must create more targeted rules built around consent, intent, distribution and provider conduct.
For AI companies, the case is a warning that written policies may no longer be enough. For lawmakers, it is a reminder that even regulations addressing deeply harmful technology must be drafted with precision. For victims, the outcome could determine whether legal protection begins only after an image spreads or before the technology is allowed to create it.
