Minnesota Governor Tim Walz has signed House File 1606, a first-in-the-nation law targeting AI nudification tools that can create or distribute nonconsensual intimate images, including images that can be used to generate child sexual abuse material, with the law set to take effect on August 1, 2026.
The law is aimed at the software layer, not just the person who uploads a bad image. It prohibits websites, apps, and other services from allowing users to create or download fake intimate images of identifiable people, and it also bans advertising or promoting those services unless significant human skill is required to create the images. That makes it materially different from older CSAM statutes, which focused on the content itself and the person who possessed or distributed it. Minnesota is now trying to reach the toolchain, which is exactly where AI abuse becomes hard to police with ordinary criminal law.
House File 1606 passed the Minnesota House 132-1 and the Senate 65-0 before Walz signed it, which tells you how broad the political consensus was around the issue. The enforcement design is also unusually aggressive. A person depicted in a violating image can sue for damages, including up to three times actual damages, punitive damages, attorneys fees, and other relief, while the Minnesota attorney general can seek civil penalties of up to $500,000 per violation. Money from penalties will flow to the general fund and then into grants for organizations helping victims of sexual assault, domestic violence, child abuse, and general crime.
The legal novelty is less about inventing a brand-new category of harm and more about deciding that the existing law was not enough for synthetic abuse at scale. Minnesota already amended its CSAM law in 2025 to cover computer-generated images depicting an indistinguishable minor engaged in sexual conduct. HF 1606 goes further by targeting the services that make such images easier to create or distribute, even before a victim has to prove a specific criminal act involving possession or sharing. That is an important distinction. Traditional CSAM law punishes the material and the offender. This law also puts pressure on the platform or tool provider that knowingly lets the material be generated in the first place.
For SF readers, this is the point where AI regulation becomes a real product and compliance problem. If one state can create a criminal and civil framework around nudification tools, then startups building generative image products, hosting layers, app stores, and open-source model tooling now face a fragmented compliance map. A company that sells image generation APIs in one state may need to prove it can block nonconsensual intimate-image outputs there, while a different state could impose a different standard entirely. That is manageable for a large platform with a legal team. It is more painful for a small startup with a consumer product and a thin trust-and-safety budget.
The law also sharpens the question of who lawmakers are really targeting. The answer is not just perpetrators. It is also the distribution channel and, indirectly, the underlying AI tool. If a service lets users create fake sexual images of identifiable people, that service is now in the enforcement path. If the service markets itself around "nudification" or easy deepfake generation, the advertising ban creates another liability layer. That is likely to matter most for companies that have tried to position themselves as generic content tools while knowing that one of the easiest abuse cases is intimate-image generation. Those firms will now have to decide whether to invest in stronger guardrails, restrict access by geography, or avoid the use case entirely.
The broader policy signal is that lawmakers are no longer content with broad AI safety language. They are moving toward concrete criminal and civil rules for specific abuse paths. That is likely to expand. Other states will study Minnesota's model, especially because it ties together victims' remedies, attorney general enforcement, and platform obligations in a single law. The result is less a clean federal standard than a patchwork of state-by-state obligations that AI companies will need to track carefully. For model companies and image-generation startups, the risk is not only that the prohibited content exists. It is that their product architecture, moderation filters, and distribution strategy may now determine whether they are viewed as a publisher, a tool maker, or an enabler of criminal misuse.
That is why this law matters beyond the obvious moral outrage around the subject. It shows that AI abuse is moving from hypothetical concern to enforceable state law. And once regulators decide they can regulate the tool as well as the act, the compliance burden lands not just on bad actors, but on every startup building anything that can create convincing synthetic media.
Also read: Baidu's Kunlunxin IPO shows China's AI chip race is now a capital markets story • OpenAI is winding down fine-tuning and that changes the startup playbook • SpaceX's $55 billion Terafab bet shows what vertical AI integration looks like at full scale
