Minnesota Fires Back at xAI: Defends First-in-the-Nation AI Nudification Ban

Minnesota has formally answered Elon Musk’s xAI in federal court, defending its first-in-the-nation ban on AI “nudification” technology and accusing the company’s Grok Imagine image generator of creating “an unparalleled marketplace for digital sexual violence.” The filing, submitted Friday by Minnesota Attorney General Keith Ellison, sets the stage for a pivotal hearing on Wednesday that could shape how every U.S. state regulates AI-generated sexual imagery.

What the filing says

At the center of the dispute is HF-1606, a Minnesota law enacted in April and effective since August 1 that prohibits website operators, software developers, and platforms from allowing users to create realistic images showing intimate body parts that did not appear in the original photo of an identifiable person. Violations carry civil penalties of up to $500,000 per image, and victims can seek damages.

In its Friday response, Ellison’s office argued that xAI is unlikely to succeed on the merits of its constitutional challenge and has failed to demonstrate that enforcing the law would cause the company irreparable harm — the two key bars a plaintiff must clear to win an injunction.

“With Grok Imagine, X.AI has created an unparalleled marketplace for digital sexual violence that poses virtually no barrier to entry,” Ellison wrote. “That marketplace, coupled with the functional capabilities of Grok Imagine, would doom Minnesota from the start if it could not direct its laws to the technology that makes digital sexual victimization possible in the first place.”

The core legal question, as both sides frame it, is whether nudification software is “speech” protected by the First Amendment or simply a “tool” that states may regulate like any other dangerous product.

The backstory: a 132-1 vote and a man with 80 victims

Minnesota lawmakers passed the measure by overwhelming, bipartisan margins — 132-1 in the House and 65-0 in the Senate — after hearings that included testimony about a man who used ordinary social media photos to generate sexualized images of more than 80 women he knew personally. State Senator Maye Quade, who led the push, said at the time: “Today, we led the nation protecting women, children and everyone in public life from the harm caused by AI nudification technology.”

xAI sued Ellison in late July, days before the law took effect, arguing that HF-1606 imposes “an overbroad, content-based ban on free speech and the tools of visual expression.” The company’s complaint stresses that the statute lacks a safe harbor: “There is no safe harbor for good-faith efforts of the provider of general-purpose AI creative tools to avoid harms. Liability attaches even if the depicted persons consented—or created the image themselves—and even if the image is never shared.”

xAI also contends the law is so broadly written it could sweep in protected images, including shirtless men, swimmers, and political satire — echoing First Amendment concerns raised by civil-liberties groups about similar bills in other states.

The company scored no early relief: on July 31, a federal judge denied xAI’s request for a temporary restraining order, allowing the ban to take effect on August 1 as scheduled. The case now moves toward Wednesday’s hearing before the same judge.

The Grok problem

Minnesota’s defense leans heavily on Grok’s documented history with explicit deepfakes. In January, a watchdog group estimated that Grok generated more than 23,000 sexualized images of children over just 11 days, prompting investigations in several countries. In March, three minors in California joined a federal class action alleging that Grok was used to turn their photos into AI-generated child sexual abuse material.

xAI says it has already acted, noting that it prohibits Grok Imagine users from generating nude or sexualized images of people without consent, has suspended more than 50,000 accounts, and filed more than 70,000 reports with the National Center for Missing and Exploited Children in 2026 alone. The company argues the law effectively demands the impossible — perfect content control — while punishing even good-faith actors.

Why this case matters far beyond Minnesota

The showdown has become a national test case for a fast-growing category of state AI legislation. Roughly two dozen states have passed some form of deepfake or AI imagery law, but Minnesota’s HF-1606 is the first to target the nudification capability itself — the tool, not just specific outputs. If the law survives First Amendment scrutiny, expect a wave of copycat legislation. If it falls, states will be forced back to narrower approaches that regulate distribution or intent rather than the underlying technology.

The case also lands amid a broader reckoning over AI safety and liability. xAI — now owned by SpaceX — has positioned itself as the “free speech” AI lab, and a win here would cement that stance in case law while undermining the legal foundation of similar bans. A loss, meanwhile, would signal that courts are willing to treat generative AI tools as regulatable products when the harms are severe enough.

Precedent cuts both ways. Courts have blocked state deepfake laws (including California’s) on First Amendment grounds when they reached political speech or satire. But regulators have historically prevailed when they can show a law targets a tool’s capacity for harm rather than a particular viewpoint — the framework Minnesota is explicitly invoking.

What happens next

Wednesday’s hearing will likely determine whether the law remains in force while the litigation proceeds. Whatever the judge decides, an appeal is nearly certain, and the eventual ruling will be parsed by statehouses, AI companies, and First Amendment scholars across the country.

For AI developers, the practical stakes are blunt: a sustained HF-1606 means image generators operating in the U.S. need consent verification, age assurance, and nudification blockers as baseline features — not optional extras. For everyone else, it is the first real answer to a question that has loomed since image models went mainstream: can the law move as fast as the technology it is trying to contain?