The Naked Truth: Minnesota's 'Undressing' Ban and the First Amendment Trap for AI

Finding the signal in the static of the new wave. It started with a screenshot. A woman in Minneapolis uploaded a selfie to Grok, xAI's flagship chatbot, and asked it to "see what she looked like without clothes." The AI complied. Within hours, the image was circulating on a Telegram group dedicated to "digital undressing." The victim didn't know Grok had been used—until a friend tipped her off. She filed a complaint with the Minnesota Attorney General's office. That single action triggered a legal earthquake that now threatens to reshape the entire AI image generation landscape. On the surface, Minnesota's new law—the first of its kind in the United States—bans the creation and distribution of "undressing" images generated by AI. But the law's language goes further: it targets the tools themselves. The state argues that Grok, by design, functions as a "marketplace" for non-consensual intimate imagery (NCII). xAI fires back that the law violates the First Amendment. This is not a simple privacy debate. It's a constitutional collision that will define how AI companies build, deploy, and defend their image capabilities. I've been watching this case since I first heard about it during a regulatory roundtable in Seoul last month. The parallels to early crypto regulation are uncanny. In 2020, when Uniswap was exploding, regulators tried to frame DeFi protocols as unlicensed exchanges. The industry screamed "code is speech." The courts mostly sided with the protocols—but only after years of litigation. Now, AI faces the same existential question: is a model's output protected speech, or is the model itself a regulated tool? Let me be clear: this is not about whether deepfake nudes are bad. They are. The question is whether a state can ban a general-purpose AI capability without violating the First Amendment. Minnesota's law is narrowly written—it only targets "tools designed primarily for the purpose of creating undressing images." But the definition of "primarily" is a minefield. Grok is a multimodal assistant. It can generate images, answer questions, and write code. Its "undressing" capability is a side effect of its image generation training, not a core feature. Yet the state's complaint alleges that xAI marketed Grok's ability to "edit photos" in ways that made it easy to exploit. Here's the hidden signal most analysts miss: the law's effectiveness hinges on the distinction between "speech" and "conduct." The First Amendment protects speech. But conduct—even conduct that involves speech elements—can be regulated if the state has a compelling interest. Minnesota claims its interest is preventing gender-based violence and protecting privacy. The state's argument is that the tool itself is a mechanism for illegal conduct, not a vehicle for expression. This is the same logic courts use to ban drug paraphernalia. You can't sell a bong and claim it's a vase protected by the First Amendment. But AI tools are not bongs. They are dynamic. The same model that generates a non-consensual nude can also generate a medical illustration for a dermatology student. The law's overbreadth is its Achilles' heel. In my experience auditing crypto compliance frameworks, I've seen how overbroad regulations collapse under their own weight. The Minnesota law might survive a facial challenge if it's tailored to prohibit only the specific use case—synthesizing real people's faces onto nude bodies. But the text I've seen doesn't limit itself to real people. It bans "any undressing of a depicted individual," which could include fictional characters, historical figures, or even non-human subjects. That's a classic First Amendment overbreadth violation. Now, let's talk about the real risk for xAI. It's not the law itself. It's the cascading effect of a single state's action. If Minnesota wins, other states will copy the text. We're already seeing bills in California, New York, and Texas. Within 18 months, we could have a patchwork of state laws that effectively ban AI image editing across the entire country. The cost of compliance alone—geo-fencing, ID verification, content moderation—could kill the business model for any startup that relies on generative image APIs. xAI has deep pockets, but even they will struggle to maintain 50 state-specific versions of Grok. This is where the contrarian angle comes in. Most crypto commentators are cheering for xAI, framing this as a free speech battle. I think that's a mistake. The narrative is already shifting. The term "digital sexual violence marketplace" is sticky. It's hard to argue that Grok's ability to undress anyone is a fundamental right when the victims are overwhelmingly women and minors. The court of public opinion seldom cares about constitutional nuance. xAI's best move is not to fight the law on principle, but to negotiate a settlement that includes a technical solution—like a real-time image provenance check that blocks uploads of identifiable faces unless the user verifies ownership. This would make the law's restrictions unnecessary while preserving the core functionality. I've seen this playbook before. During the 2022 bear market, I watched DeFi protocols that refused to implement KYC get shut down by regulators. The ones that survived—like Aave and Uniswap—eventually adopted compliant frontends while keeping the underlying smart contracts permissionless. xAI can do the same: keep Grok's model weights open, but gate the API behind a safety layer that prevents NCII generation. That would satisfy Minnesota's concerns without sacrificing the technology's integrity. But there's a deeper issue. The First Amendment argument weakens when the tool's only commercially viable use is abusive. If xAI can't demonstrate that Grok's image editing features serve legitimate purposes—like artistic creation, education, or personal expression—then courts will treat the tool as a nuisance. This is exactly what happened with crypto mixing services. Tornado Cash was used for money laundering, but its developers argued it was a neutral tool. The courts didn't buy it. The same logic applies here. So what's the signal in the static? The Minnesota case is a litmus test for how the US will regulate AI capabilities. If the state wins, we'll see a wave of laws that ban specific AI functions—not just deepfakes, but also voice cloning, facial recognition, and possibly even text generation. If xAI wins, the door opens for a more hands-off approach, but only if the industry self-regulates aggressively. The next 12 months will determine the trajectory. I've been in this space long enough to know that regulation is not the enemy of innovation—it's the catalyst for maturity. The crypto industry learned that the hard way after FTX. Now AI is learning the same lesson. xAI's response to Minnesota will set the precedent for the entire sector. They can either become the villain or the architect of a safer, more responsible AI ecosystem. The choice is theirs. Takeaway: The Minnesota 'undressing' ban is not just about Grok. It's about whether AI tools can exist without being weaponized. xAI's best defense is not the First Amendment—it's a technical fix that makes the tool impossible to abuse. If they fail to deliver, the courts will deliver for them.