Minnesota Attorney General Keith Ellison speaks during a press conference at the State Capitol building on February 3, 2026 in St. Paul, Minnesota.
Stephen Maturen/Getty Images
Minnesota’s first-in-the-nation ban on AI nudification technology took effect August 1, after a federal judge denied Elon Musk’s artificial intelligence company a last-minute stay — and within days, three new lawsuits from a UK lawmaker, an Arkansas family, and five children joined the growing legal front against Grok Imagine. The real constitutional test, however, is still weeks away: the ruling that let the law stand resolved nothing on the merits, and a preliminary injunction hearing scheduled for August 19 will mark the first time any court examines whether HF 1606’s strict-liability structure can survive First Amendment scrutiny.
Judge Denies TRO on Procedural Grounds: Law Now in Effect
U.S. District Judge Donovan W. Frank denied xAI’s TRO request on July 31, allowing HF 1606 to take effect on schedule the following day. The ruling did not touch the constitutional arguments xAI has raised — that the law is a content-based restriction and imposes strict liability without any safe harbor for platforms that prohibit and actively filter nudification. Instead, Judge Frank focused entirely on timing.
“xAI filed the motion on July 29, 2026, nearly three months after the law was signed, and only three days before the law is set to take effect,” the judge wrote. “Such a delay in bringing the action and the motion suggests that harm is not immediate.” The full quote is documented in the Minnesota AG’s July 31 press release.
xAI filed its lawsuit on July 27, 2026 — nearly three months after Gov. Tim Walz signed HF 1606 into law on May 6. The company sought an emergency restraining order two days later. Minnesota Attorney General Keith Ellison, in his response, argued that xAI’s voluntary delay “should not create a manufactured emergency for the attorney general or the court” — a position Judge Frank accepted without reaching the First Amendment questions.
Rather than dismissing the motion outright, Frank converted it into a preliminary injunction request and set a hearing for August 19 at 9:30 a.m. ET at the Warren E. Burger Federal Building in St. Paul. Attorney General Ellison must file his opposition by August 12; xAI’s reply brief is due August 17. The expedited August 19 briefing schedule was published by CBS Minnesota.
For now, xAI has implemented the practical consequence it warned about: the company has restricted Grok Imagine’s image-editing features for users in Minnesota, citing the law’s $500,000-per-incident civil penalty structure as leaving the company with “no practical choice” but to restrict its features in the state.
What Minnesota’s Law Actually Does — and How It Differs
HF 1606’s full legislative text prohibits anyone who owns or controls a website, application, software, or other service from allowing users to access, download, or use it to nudify an image or video of a real, identifiable person — meaning to generate or alter an image to depict intimate parts not present in the original. The law targets platform operators rather than individual users, imposes civil penalties of at least $500,000 per unlawful incident, and empowers both victims and the attorney general to sue.
The legislation passed with near-unanimous 132-to-1 House vote and a 65-to-0 Senate vote — a margin that reflects the degree to which legislators across the political spectrum view nonconsensual deepfake imagery as beyond ordinary political dispute.
What makes the law structurally distinct from most existing state NCII statutes — and constitutionally controversial — is its upstream approach. Most states have criminalized the dissemination of nonconsensual intimate imagery; Minnesota’s law targets the tool itself. An image nudified through Grok or any similar service triggers liability the moment it is generated, regardless of whether it is ever shared. That design reflects a legislative conclusion that distribution-targeting laws act too late: by the time a victim discovers that nude images of her exist and requests removal, those images may already have spread through closed messaging apps, anonymous accounts, or school networks where removal is meaningless.
What August 19 Will Actually Decide
The TRO denial was a narrow procedural loss for xAI — it means the law is currently in force, but it means nothing about whether the law is constitutional. That question reaches its first genuine judicial test at the August 19 preliminary injunction hearing.
To win a preliminary injunction, xAI must demonstrate a likelihood of success on the constitutional merits — not merely that filing was timely. The arguments it will make are the same ones it raised in its 39-page federal complaint and that Judge Frank has not yet assessed.
First, that HF 1606 is a content-based restriction under the Supreme Court’s 2015 ruling in Reed v. Town of Gilbert — because enforcing the law requires examining what an image depicts — and must therefore survive strict scrutiny, which demands narrow tailoring through the least restrictive available means.
Second, that the law’s strict liability structure, which imposes $500,000-per-incident exposure on platforms regardless of their policies, filters, or good-faith compliance efforts, is not narrowly tailored. xAI argues — and the technical reality confirms — that modern AI image editors use the same diffusion model inpainting architecture for all editing tasks, and no content filter achieves a zero false-negative rate for all possible user prompts. A law holding platforms strictly liable for every user who bypasses filters that blocked thousands of prior attempts cannot, xAI argues, survive the narrow-tailoring requirement.
xAI concedes Minnesota’s interest in preventing nonconsensual nude imagery is compelling. The dispute at August 19 will be entirely about narrow tailoring — whether a law with no scienter and no safe harbor for compliant platforms is the least restrictive means of achieving that goal.
Minnesota has until August 12 to argue it is. The outcome of that hearing — not the TRO denial — is what will determine whether every state legislature in the country can use HF 1606 as a model.
Grok’s Deepfake Record Hangs Over the Litigation
xAI brought the lawsuit as a constitutional-law challenge. The documented record of Grok’s behavior has made it an awkward plaintiff.
Between December 29, 2025 and January 8, 2026 — an 11-day window following Elon Musk’s public promotion of Grok’s image-editing capability on X — researchers at the Center for Countering Digital Hate estimated that Grok generated 3 million sexualized images, including around 23,000 that appeared to depict children.
Rather than disabling the feature after the outcry, the company restricted explicit image generation to paying subscribers and marketed “Spicy Mode” as a premium benefit, according to allegations in the Lieff Cabraser class action complaint.
Grok was still generating sexualized images of people as recently as April 2026, despite restrictions introduced in January, according to NBC News April reporting.
Attorney General Ellison, in his July 31 public statement praising the TRO denial, was direct about the stakes: “These nudification apps, including Grok Imagine, have been used to generate child sexual abuse materials and harass people in the vilest ways imaginable.”
SpaceX — xAI’s parent company following the February 2026 merger — has earmarked more than $530 million in its IPO filing for potential Grok-related litigation losses, as documented in Memeburn’s comprehensive Grok tracker.
Three New Suits Filed in the Same Week
In the same week that HF 1606 took effect, three new legal actions targeting Grok Imagine were filed or escalated — the precise development that Archewell Philanthropies cited in its August 2 statement.
The UK High Court claim: Labour MP Jess Asato, who represents Lowestoft, filed a claim at the UK High Court on June 3, 2026, alleging that Grok was used to produce nonconsensual deepfake images of her, including content allegedly depicting her being sexually assaulted. The claim alleges violations of UK data protection law and misuse of private information, details of which are included in Memeburn’s 2026 Grok lawsuit tracker.
The Arkansas family lawsuit: Parents identified as John and Jane Roe filed a federal CSAM lawsuit in Little Rock on July 23, 2026, on behalf of their 10-year-old daughter. The complaint alleges that a Bentonville photographer — Russell Bloodworth, who pleaded not guilty to more than 200 criminal charges — used Grok to generate child sexual abuse material from photographs he had taken of the child. The lawsuit describes xAI as having made the generation of sexual content of real people “a feature, not a bug” — noting that “every other major AI image and video provider wholly refuses to generate sexual content of real people.”
The five-child class action: A class action originally filed by three Tennessee minors has been amended to add five plaintiffs across multiple states, with Stability AI added as a defendant. In one case documented in the expanded complaint, a Wyoming man’s devices held approximately 7,000 AI-generated images and videos of his stepdaughter, all produced using Grok.
xAI disputes all of the allegations in the civil lawsuits, points to its terms of service prohibiting nonconsensual intimate imagery, and has cited more than 52,000 account suspensions and more than 73,000 reports filed with the National Center for Missing & Exploited Children in 2026. The amended complaint’s NCMEC allegations contend, however, that the National Center for Missing and Exploited Children determined that 90 percent of xAI’s CyberTipline reports were non-actionable by law enforcement because the company declined to include identifying user information.
Archewell Philanthropies Enters the Fight
On August 2, Prince Harry and Meghan, Duke and Duchess of Sussex, published a statement through their charitable organization, Archewell Philanthropies, praising Minnesota’s law and directly criticizing xAI’s effort to block it. The statement, headlined “Can we all agree technology should not enable predators to target children?”, called the legislation “an example of leadership fit for the digital age: a bipartisan decision to stop technology from enabling predatory behavior disproportionately targeting women and children,” as reported by GB News from Sussex.com.
Without naming Musk, the statement’s sharpest language targeted xAI directly: “Big Tech has taken shocking steps to limit restrictions on or accountability for these objectively predatory AI product features.” It singled out “the trillionaire leader of the tech company responsible for X’s Grok” for having sued the state of Minnesota to stop the law from taking effect.
The statement cited UN Women’s deepfake research indicating that 90 to 95 percent of all online deepfakes are nonconsensual pornographic images, with approximately 90 percent depicting women.
Archewell Philanthropies’ involvement marks the entry of high-profile international advocacy into what had been primarily a domestic US legal dispute. The organization has previously worked on digital safety and has a Parents’ Network supporting families affected by online harm. The Sussex.com statement did not include a response from xAI or from Musk.
A Broader Legal Battle Still Taking Shape
The Minnesota case sits within a larger regulatory and litigation landscape that xAI now faces across at least nine jurisdictions. Regulators in the European Union, the United Kingdom (Ofcom), California, Canada (the Office of the Privacy Commissioner), India, Indonesia, Malaysia, and Australia have all opened investigations into Grok’s outputs or taken enforcement actions, as documented in Memeburn’s global regulatory tracker.
The Trump administration has separately indicated interest in challenging state AI laws and establishing a single national regulatory framework — a posture that could eventually preempt Minnesota’s approach. The Department of Justice intervened in xAI’s Colorado AI Act challenge in April 2026, in what was DOJ’s first intervention against a state AI law under President Trump’s Executive Order 14365.
Minnesota’s law, if it survives the August 19 hearing, would represent the most aggressive state platform-liability model in the country. Laws targeting nudification platform operators are also in force in Pennsylvania, North Dakota, and California, though none imposes per-incident civil penalties at the $500,000 level that HF 1606 does.
The law also creates a question the August 19 hearing may not fully resolve: whether the $500,000-per-incident structure combined with no scienter and no safe harbor for compliant platforms chills general-purpose image editing across the entire AI sector — or whether it can be narrowly construed to apply only to platforms specifically designed or marketed for nudification. That interpretive question, more than the headline penalty figure, is what the national legal community will be watching when Judge Frank takes the bench on August 19.
Frequently Asked QuestionsWhat did the July 31 court ruling actually decide — and what didn’t it decide?
Judge Frank’s July 31 ruling denied xAI’s request for a temporary restraining order on procedural grounds: the company waited nearly three months after HF 1606 was signed before seeking emergency relief, and the court concluded that delay undermined the claim of irreparable harm that a TRO requires. The ruling did not assess whether HF 1606 is constitutional. The question of whether the law’s strict-liability-without-scienter structure survives First Amendment scrutiny will be addressed for the first time at the August 19 preliminary injunction hearing. A full analysis appears in The Next Web’s TRO ruling coverage.
What does it mean that xAI is now restricting Grok Imagine for Minnesota users?
xAI stated in its lawsuit that the $500,000-per-incident civil penalty structure left it with no practical choice but to restrict Grok Imagine’s image-editing features for users in Minnesota once the law took effect, as detailed in eGamers’ $500,000 penalty analysis. The company did not specify which precise features are being limited, but the restriction reflects the law’s design: Minnesota’s platform liability approach works by making the financial cost of enabling nudification prohibitively high, even for a platform that prohibits it in its terms of service. Users in other states are not currently subject to equivalent restrictions.
Who else is suing xAI over Grok deepfakes, and what are they claiming?
As of early August 2026, xAI faces at least six major legal actions across the US and UK related to Grok’s image generation capabilities, as tracked in Memeburn’s 2026 Grok lawsuit roundup: (1) the Lieff Cabraser class action in California, now representing five children from multiple states, including one child whose stepfather used Grok to generate approximately 7,000 explicit images from a single childhood photograph; (2) a federal lawsuit filed July 23 in Little Rock by Arkansas parents of a 10-year-old whose photos were used to generate CSAM by a Bentonville photographer facing 200-plus criminal charges; (3) a UK High Court claim by Labour MP Jess Asato, alleging deepfake images of her — including content allegedly depicting sexual assault — were created using Grok; (4) the Baltimore City consumer protection lawsuit; and (5) the Minnesota AG enforcement action under HF 1606. Each case raises related but legally distinct theories ranging from CSAM production liability to data protection violations to consumer fraud.
Why does the August 19 hearing matter more than the July 31 TRO denial?
The July 31 TRO denial was a procedural ruling on timing — it says nothing about whether Minnesota’s law is constitutional. The August 19 preliminary injunction hearing is the first occasion on which Judge Frank will evaluate xAI’s core legal argument: that HF 1606 is a content-based restriction requiring strict scrutiny under Reed v. Town of Gilbert, and that its strict-liability-without-scienter structure fails the narrow-tailoring requirement because it holds compliant platforms equally liable with bad actors. If xAI wins a preliminary injunction on August 19, the law goes back on hold while the full lawsuit proceeds. If it loses, the law stays in force. Either outcome will signal to legislators in California, New York, and other states whether Minnesota’s model is constitutionally viable — making the hearing the most consequential AI platform-liability ruling since HF 1606 passed 197 to 1, as analyzed in IBTimes UK’s preliminary injunction coverage.