xAI is suing Minnesota Attorney General Keith Ellison over a law that bans “nudification” apps, arguing the statute violates the First Amendment and forces the company to restrict Grok’s image-editing features.

The timing is notable. Minnesota passed the law back in May, just months after Grok flooded the internet with millions of sexually explicit deepfakes in January, including images of minors. Now xAI is arguing the law goes too far.

According to the company’s complaint, the statute’s punitive provisions leave xAI with “no practical choice but to restrict Grok Imagine’s image-editing features in various ways.” The law broadly targets apps and services that create non-consensual nude or sexually explicit images of real people.

xAI’s legal theory is straightforward: the law is an unconstitutional restriction on speech. The company isn’t arguing that creating deepfake nudes of real people without consent is good or should be legal everywhere. It’s arguing that Minnesota’s specific statute is too broad and conflicts with the First Amendment’s protection of expression, even when that expression involves AI-generated imagery.

This is the core tension in regulating AI image generation. States want to crack down on harmful deepfakes, particularly non-consensual sexual imagery. But when they write laws broad enough to capture all the bad uses, they risk sweeping in legitimate uses of image-editing technology. xAI is betting a court will find Minnesota crossed that line.

The company has some precedent to work with. Courts have repeatedly struck down state laws that restrict speech too broadly, even when the underlying goal is preventing harm. The question is whether creating synthetic nude images of real people falls within First Amendment protection, or whether it’s more like defamation or invasion of privacy, where the law can impose limits.

Minnesota’s law is part of a wave of state legislation targeting AI-generated sexual imagery. Multiple states have passed similar statutes in the past year, responding to the explosion of deepfake technology that can create realistic fake nudes from clothed photos. The technology has been weaponized against women, celebrities, and increasingly, ordinary people who find fake nude images of themselves circulating online.

But xAI’s track record here complicates its position. The January incident wasn’t a small-scale problem. Grok’s image generator produced millions of explicit deepfakes before the company implemented restrictions. Users created fake nudes of celebrities, politicians, and others, flooding social media with non-consensual sexual imagery. Some of the images depicted minors.

The company did eventually add content moderation. But it only acted after the problem became a public crisis. Now it’s arguing in court that Minnesota can’t force it to maintain those kinds of restrictions.

xAI’s complaint doesn’t dispute that some uses of nudification technology are harmful. The company’s argument is narrower: Minnesota’s law is too broad and punishes too much protected speech along with the unprotected stuff. If xAI is right, the state will need to write a more carefully tailored statute that targets specific harms without sweeping in legitimate image editing.

If xAI loses, the company will have to comply with Minnesota’s restrictions or block Grok Imagine from users in the state. That could set a precedent for other states to pass similar laws, creating a patchwork of regulations that AI companies would need to navigate.

The lawsuit also raises questions about what happens when AI companies face conflicting state laws. If Minnesota can ban nudification features and another state can’t, does xAI build different versions of Grok for different jurisdictions? Or does it apply the most restrictive standard everywhere?

These aren’t hypothetical problems. They’re the practical reality of trying to regulate AI tools that can be used for both legitimate and harmful purposes, and that are accessible from anywhere with an internet connection.

The Minnesota Attorney General’s office hasn’t filed a response yet, so we don’t know how the state will defend the law. But the likely argument is straightforward: creating fake nude images of real people without their consent isn’t protected speech, and the state has a legitimate interest in preventing that harm. The law isn’t targeting general image editing. It’s targeting a specific use case that inflicts real damage on real people.

Both sides have reasonable arguments. The First Amendment does protect a lot of offensive speech, including artistic uses of AI that might make people uncomfortable. But it doesn’t protect everything, and laws against things like defamation, harassment, and invasion of privacy survive constitutional scrutiny.

The case will likely turn on how courts classify AI-generated nudification. Is it more like photoshopping someone into a fake scenario (which is generally legal, even if offensive)? Or is it more like revenge porn (which many states have successfully criminalized)?

xAI is making a bet that courts will see this as a speech issue first and a harm issue second. Minnesota is betting the opposite. Whoever wins will help define the legal boundaries for AI image generation going forward.

The lawsuit is pending in federal court. No hearing date has been set.