Minnesota's "Nudification" Ban Repeats Unconstitutional Errors of Early Revenge Porn Bills
Minnesota’s new law banning AI powered “nudification” tools is facing a constitutional challenge over concerns that it restricts substantially more speech than necessary to address nonconsensual sexual deepfakes. Constitutional law professor Josh Blackman argues that although preventing the creation and distribution of nonconsensual intimate imagery is a legitimate government objective, Minnesota’s legislation repeats problems found in early revenge porn laws by failing to require knowledge or intent before imposing liability.
The law prohibits operators of websites, applications, software and other services from allowing users to access or use their services to “nudify” an image or video. Nudification is broadly defined to include altering or generating an image depicting an intimate body part that was not present in the original image of an identifiable person.
Blackman argues that the central constitutional problem is the absence of a scienter requirement. Under the statute, a service provider could potentially face liability based solely on what a user manages to generate, regardless of whether the provider knew about, intended or could reasonably have prevented the activity. The legislation also does not distinguish between consensual and nonconsensual uses. Consequently, someone altering an image of themselves could technically fall within the same prohibition as a stranger creating an unauthorized nude image of another person.
The article compares this approach with earlier laws targeting revenge porn. Initial legislation in several states effectively imposed strict liability for distributing intimate images, raising significant First Amendment concerns. Those laws were subsequently narrowed by introducing requirements that defendants knowingly or intentionally distribute private sexual material. Minnesota’s own Supreme Court upheld the state’s revenge porn law partly because its specific intent requirement excluded negligent, accidental and reckless conduct.
Another concern is that Minnesota’s nudification statute contains no meaningful exceptions for artistic, political, medical, educational or other potentially protected uses. According to Blackman, this makes the prohibition substantially broader than necessary and potentially captures legitimate expression alongside abusive deepfake content.
xAI, the company behind Grok, has challenged the law in federal court. A district court denied its request for a temporary restraining order, and proceedings concerning a preliminary injunction are continuing. xAI argues that Minnesota could legitimately prohibit intentionally generated nonconsensual nude images while still protecting constitutionally protected uses of generative AI.
xAI also maintains that it already prohibits users from generating nude or sexualized images of people without their consent and employs technological safeguards designed to prevent such activity. The company has even pursued legal action against users accused of circumventing those safeguards. Its constitutional argument therefore does not dispute the harm associated with abusive nudification, but instead focuses on whether Minnesota has designed its law narrowly enough to address that harm without restricting protected speech.
Minnesota has argued that the statute should receive intermediate constitutional scrutiny and that a scienter requirement is therefore unnecessary. Blackman rejects this position, arguing that the law is inherently content based because liability depends directly on what an AI generated image depicts. In his view, that means stronger First Amendment scrutiny and narrow tailoring are required.
Overall, the article argues that governments can and should address malicious AI generated intimate imagery, but legislation must distinguish intentional abuse from legitimate uses. Blackman concludes that Minnesota could create a narrower law targeting deliberate nonconsensual nudification, but that the current statute is constitutionally vulnerable because of its broad scope and lack of an intent requirement.
