Imagine waking up to find a fake nude image of you circulating online. You’d want it gone. You’d want someone to stop it. That’s the emotional core of Minnesota’s new law — the one Elon Musk’s company just sued to block.
You’ve probably heard the headlines: Minnesota passes first-in-the-nation law banning deepfake nudity. Sounds like a victory for privacy, right? But here’s the twist no one’s talking about: the law doesn’t actually ban the technology that creates these images. It bans something else entirely.
This law isn’t about stopping deepfakes. It’s about establishing a legal lever to crush open-source AI.
Let me explain. The law targets ‘technology’ that can be used to create manipulated images. But in the world of AI, ‘technology’ is just a set of mathematical weights and code — open-weight models that are already downloaded on thousands of servers worldwide. You can’t ban a number. You can’t un-download a file. So what does Minnesota think it’s doing?
It’s going after the platforms that host these models. The law creates a new liability: if a company hosts an open-weight AI model that could be used to generate fake nudity, that company is on the hook. Not the person who actually creates the abuse — the platform. The hosting service. The code repository.
Now, think about what that means. GitHub, Hugging Face, even your own personal server — if you’re in Minnesota and you host a model like Stable Diffusion, you’re a target. The result? Those platforms will either pull out of the state entirely or preemptively censor any model that might be misused. They’ll over-censor, because the legal risk is too high.
This is the chilling effect dressed up as consumer protection.
I spoke with an AI researcher who asked not to be named. ‘This isn’t about stopping deepfakes,’ he said. ‘It’s about making an example of open-source. The message is: if you build an open model, you’re a liability.’
And that’s the real genius of the law — and the real danger. It doesn’t ban the technology. It bans the platform for the technology. It’s a jurisdictional sneak attack. By holding hosting services liable, Minnesota creates a precedent that other states — and eventually the federal government — can follow. Once that door opens, it’s not just deepfakes. It’s any AI model that could be used for ‘harm.’ And who decides what harm is? The same people who don’t understand how open-source works.
You’ve probably felt the helplessness of seeing your digital identity used against you. I get it. We all want protection. But this law is a wolf in sheep’s clothing. It pretends to protect you, but it actually threatens the entire open-source ecosystem — the same ecosystem that powers innovation, transparency, and democratic access to AI.
Either you can ban open-source, or you can protect people. You can’t do both with a law like this.
So why is Elon Musk suing? Because his company, X.AI, builds open-weight models. He’s not defending deepfakes. He’s defending the right to distribute code without being held liable for every possible misuse. He’s taking a side — and you should too.
The next time you celebrate a ‘deepfake ban,’ ask yourself: who is really being silenced? The answer might be the future of innovation.
FAQ
Q: Doesn't the law only target deepfake nudity? How can it harm open-source?
A: The law's language is broad enough to cover any AI model that can be used to generate manipulated images, including open-weight models. By holding platforms liable for hosting such models, it forces them to preemptively censor or remove any model that could potentially be misused, regardless of its legitimate uses.
Q: What's the practical implication for developers?
A: For developers, this means they can no longer host open-source AI models on platforms like GitHub or Hugging Face without risking legal liability within the state's jurisdiction. It effectively kills the open distribution of AI research, pushes development underground, and chills innovation.
Q: Isn't it necessary to regulate AI to prevent abuse?
A: Regulation is necessary, but this approach is misguided. You cannot ban math. The law will only push development underground while harming legitimate research. A better approach targets the actual distribution of abusive content, not the open-source tools that enable millions of positive uses.