Court Upholds Minnesota's Nudify App Ban as xAI Lawsuit Fails at First Hurdle

The first US state law specifically targeting AI-powered tools for generating non-consensual explicit imagery is now in effect, after a federal judge declined to halt it on the eve of its enforcement date. The ruling is the first significant legal test of a new wave of state-level AI content regulations—and the outcome will shape how every major image-generation platform operates in the United States.

Minnesota’s HF 1606, signed into law during the spring 2026 legislative session, imposes a $500,000 civil penalty per instance against developers whose tools are used to generate non-consensual explicit deepfakes of real people. The law passed the Minnesota House 132-1 and the Minnesota Senate 65-0, a degree of bipartisan consensus that is rare for technology legislation.

The Emergency Motion

Elon Musk’s AI company xAI filed suit against Minnesota Attorney General Keith Ellison in July, seeking a preliminary injunction to prevent the law from taking effect on August 1. The company argued HF 1606 violated First Amendment protections as an “overbroad, content-based restriction on free speech and the tools of visual expression.”

Judge Donovan Frank denied the motion on July 31, hours before the law came into force. In his ruling, Frank noted that xAI had waited nearly three months after the bill was signed before filing its emergency motion—and submitted the filing just three days before the deadline. The delay, Frank wrote, “fundamentally undercuts xAI’s claims of urgent harm.” An emergency motion requires demonstrating imminent, irreparable injury; that is difficult to establish when a plaintiff waited most of the law’s pre-enforcement window before seeking relief.

The underlying lawsuit is not over. xAI is pressing its First Amendment challenge in full proceedings, and the company’s legal arguments remain live before the court.

xAI’s Legal Arguments

The company’s core challenge targets the law’s definitional scope. HF 1606 defines prohibited content through the concept of “intimate part,” and xAI argues the definition is drawn broad enough to capture clearly legal material—photographs of people in swimwear, shirtless individuals, or other images that have no connection to the non-consensual pornography the law is ostensibly designed to prevent.

The company also argues the law effectively imposes strict liability on developers regardless of whether they had any knowledge of a user’s intent to generate harmful content. “The statute sweeps up far more than the harmful conduct it claims to target,” xAI told the court.

xAI noted that its Grok Imagine product already prohibits the generation of explicit imagery and that the company has pursued litigation against users who bypassed those safeguards. The argument is essentially that the company does not need this law because it already enforces the relevant policies voluntarily—and that the law’s collateral damage on legitimate expression makes it constitutionally infirm regardless.

The Counterarguments

Minnesota lawmakers and Attorney General Ellison have defended the statute on both substantive and procedural grounds.

On substance: the legislative record shows HF 1606 was drafted specifically in response to documented harm—instances where Grok Imagine was used to generate non-consensual explicit images of private individuals, and cases where similar tools were deployed against minors. The bill’s sponsors argue the $500,000 per-instance figure is calibrated to create genuine deterrence rather than a cost-of-doing-business situation where developers absorb smaller fines as acceptable overhead.

On procedure: the bipartisan legislative record—132-1 in the House, 65-0 in the Senate—suggests the law reflects broad consensus about the harm being addressed. Courts applying First Amendment scrutiny give greater deference to legislative judgments with this kind of supermajority support.

On xAI’s specific platform: the company’s claim that it already prohibits this content is, in the legislative record, precisely the context the law was responding to. Users bypassed Grok Imagine’s own content filters, and the harm occurred anyway. The question of developer liability in that circumstance is exactly what the statute addresses.

Why This Case Matters Beyond Minnesota

HF 1606 is the first but almost certainly not the last state-level AI content law of this type. Similar legislation is advancing in at least a dozen other states. The constitutional framework being litigated in this case will set the template for how those laws are drafted, challenged, and ultimately upheld or struck down.

If xAI prevails in the underlying suit, future nudify statutes will need to incorporate narrower definitions of prohibited content, clearer scienter requirements, and explicit safe harbors for platforms that implement meaningful technical safeguards. That outcome would be a significant constraint on state legislative authority in this space, but it would also produce more precise and defensible laws.

If the injunction continues to be denied and the law survives constitutional challenge, expect image-model vendors to geofence features in Minnesota and other states that follow its lead—disabling certain generation capabilities for users in those jurisdictions. That approach would let companies comply without withdrawing products nationally, but it would also demonstrate that geographic content restriction is technically feasible, potentially reducing industry resistance to similar laws elsewhere.

The Broader Deepfake Regulatory Landscape

The Minnesota case is landing in a complicated federal environment. The DEFIANCE Act, which created a federal civil remedy for victims of non-consensual deepfakes, passed in 2024 but targets individual perpetrators rather than platform developers. HF 1606’s developer liability model is a distinct approach that the federal government has not yet adopted.

The Trump administration’s general posture on AI regulation has been permissive, and there is no current indication of federal action to preempt state laws in this area. That creates a patchwork risk for industry: companies operating nationally may face meaningfully different legal obligations in different states, driving up compliance costs and potentially incentivizing developers to consolidate development in the most permissive jurisdictions.

For xAI, the immediate stakes are financial. At $500,000 per instance, even limited enforcement exposure in a state with five million people could be substantial. The company’s own estimate of potential liability—$50 billion for 100,000 images—reflects the scale of what strict enforcement could theoretically produce. That number was presumably crafted to illustrate the law’s allegedly disproportionate scope, but it also reveals how seriously the company is taking the financial exposure.

The case will continue in the trial court. No timeline for resolution has been established.