The Justice Department filed a statement of interest on 18 August backing Elon Musk's xAI in its challenge to a Minnesota law that bans AI "nudification" tools [1]. A federal judge heard arguments on Wednesday and said he would rule soon, according to Courthouse News Service [2], which puts a federal thumb on the scale against a state content rule that product teams were already building to.
The mechanics matter more than the politics. HF 1606 took effect this month and bars anyone who owns or controls a website, app or software from letting users nudify or sexually alter images with AI, with penalties up to $500,000 [3]. In its complaint, xAI describes it as a strict-liability rule: liability turns only on whether a user managed to produce a covered image, regardless of what the platform knew or intended [4]. That is a different engineering problem from a takedown obligation. xAI says its policies already forbid generating nude or sexualised images of real people without consent, and that it has sued users who route around its blockers [5].
The DOJ filing, signed by Associate Attorney General Stanley Woodward, says Washington shares Minnesota's goal of fighting sexual abuse imagery but objects to how the statute was drafted [6]. It did not take a position on the injunction itself, arguing instead that the Minnesota law reaches further than federal law and warning that heavy state rules could slow the AI industry [7]. The department pointed to the PROTECT Act and the TAKE IT DOWN Act, signed by President Trump in 2025, which sets a notice-and-removal system giving platforms 48 hours to remove flagged images and is enforced by the Federal Trade Commission [8]. Federal law, the filing says, requires a knowing act, exempts matters of public concern, and turns on an adult's lack of consent, while HF 1606 has none of those limits and can apply when an adult makes an image of themselves [9]. The department's example: an AI image of a shirtless man in a swimming pool, which federal law would not cover but Minnesota's broad definition of "intimate part" could [10].
Minnesota's position is that after-the-fact enforcement does not work. Assistant Attorney General Janine Kimble told the court terms of service are not enough, noted that xAI is still reporting tens of thousands of users who create such images, and said tools like Grok have no real barrier to entry [11]. "The repercussions only happen after the fact," she said, because someone has to see an image, report it, locate it, and reach a user a court can touch [12]. The state also cited a figure that around 95% of deepfake images are nonconsensual sexually explicit depictions of other people [13]. xAI's attorney Robert Dunn argued the statute has no carve-outs for consent, artistic expression or parody, and could be fixed by focusing on consent and distribution plus a safe harbor for platforms acting in good faith [14].
The filing ties this to the administration's broader position, citing executive orders that call for a single national standard rather than "50 discordant state ones" [15]. The administration has separately pushed to pre-empt state AI rules in exchange for federal legislation [16].
Watch Judge Donovan Frank's ruling in X.AI LLC v. Keith Ellison [17]. Then watch whether the DOJ files the same argument against other state statutes, and whether the safe-harbor and knowledge-standard language Dunn asked for shows up in the next round of state drafting. If it does, the compliance target moves from gating at generation to notice-and-removal, and those are different roadmaps.
