Not a Fun Party Trick: AI Sexual Harassment and First Amendment Challenges

An analysis of how the rapid rise of AI deepfakes is fueling online sexual harassment

---

Published July 6th, 2026

Written by Kathy Bian

Content warning: discussions of sexual harassment and descriptions of sexual violence

The integration of AI into everyday life has been rapid and widespread. From AI tools embedded in various applications to the viral AI-generated TikToks, it’s hard to ignore how much AI is now part of our everyday experiences. AI, however, is no longer just a fun party trick. Recent debates focus on the construction of AI data centers in resource-rich areas like Wisconsin, the regulation and management of data center water use, and the ethical issues surrounding AI, including deepfakes and pornographic content. These generated deepfakes and explicit content introduce new legal challenges, set important precedents and raise concerns surrounding First Amendment protections and sexual harassment criminal laws. 

The University of Virginia Information Security defines a deepfake as “an artificial image or video (a series of images) generated by a special kind of machine learning called ‘deep’ learning.” Deep learning occurs when a machine algorithm is fed examples and learns to produce outputs similar to the examples. This type of technology can replace faces in images and videos, manipulate facial expressions, and create new speech. With the ease of access to AI deepfake tools and websites, anyone with basic computer knowledge can create and distribute deepfake content. Deepfakes are mainly used for exploitation; the vast majority of deepfake content is pornographic and disproportionately victimizes women, according to The GAO Science, Technology Assessment, and Analytics Team. A notable example of this is Elon Musk’s AI chatbot Grok. The New York Times reported that in just nine days of launching Grok on X, the chatbot posted more than 4.4 million images, and about 41% of those images contained sexualized images of women. Wired describes one of the many graphic AI-generated videos: “One photorealistic Grok video… shows a fully naked AI-generated man and woman, covered in blood across the body and face, having sex, while two other naked women dance in the background.” 

As AI technology advances, so have state and federal laws. On April 28th, 2025, the Tools to Address Known Exploitation by Immobilizing Technological Deepfakes On Websites and Networks Act, colloquially known as the TAKE IT DOWN Act, was passed bipartisanly and signed into law. The Act seeks to criminalize non-consensual intimate images (NCII) that include AI-generated images. It requires websites to remove NCII within 48 hours of a survivor’s request, protects medical officials and law enforcement who practice ethical disclosure of NCII and protects First Amendment rights as summarized by RAINN. As of August 2025, 45 states have enacted laws that criminalize AI-generated child sexual abuse material, according to Enough Abuse. Wisconsin Act 224 of 2023 ensures that the “receiving, distributing, producing, or possessing, or accessing in any way with the intent to view, obscene material that contains a depiction of a purported child” is a Class D felony, punishable by up to 25 years in prison. 

Legal challenges have arisen in courts over the boundaries between free speech and AI-generated content that may be considered obscene. Steven Anderegg from Holmen, WI, was arrested in May 2024 “for allegedly producing and distributing AI-generated images of minors engaged in sexually explicit conduct,” according to a statement produced by the U.S. Department of Justice. In February 2025, a federal judge in the Western District of Wisconsin dismissed one out of four charges, ruling that the “private production of obscenity” is protected by Stanley v. Georgia (decided in 1969). The U.S. attorneys in the case have appealed, arguing that the Stanley decision does not apply to CSAM.

  In future cases, courts should weigh protecting the lives of those affected by sexualized AI-generated content more heavily than prioritizing First Amendment rights by applying the obscenity standard, or prurient interest, to sexually explicit AI content, known as the Miller test. Miller v. California decided in 1971, established that content is “obscene” if (1) whether the average person applying contemporary community standards would find the work, taken as a whole, appeals to the prurient interest; (2) whether the work depicts or describes, in a patently offensive way, sexual conduct specifically defined by the applicable state law; and (3) whether the work, taken as a whole, lacks serious literary, artistic, political or scientific value. On the face of it, AI-CSAM and other AI sexual content would meet these three criteria: the average person would find child pornography, taken as a whole, appealing to the prurient interest, it is considered offensive sexual content regulated and criminalized by state laws, and CSAM lacks any literary, artistic, political, or scientific value. Even if it can be argued that there is no “real child” involved, AI-generated sexual content can still inflict irreparable harm, such as psychological trauma and the normalization of sexual abuse towards women and children. 

The rapid expansion of AI technology exposes gaps in First Amendment jurisprudence that courts should not ignore. While freedom of speech is a cornerstone of American democracy, it is not absolute, especially at the expense of victims of sexual assault. Prioritizing the protection of individuals from irreparable harms caused by sexualized AI content ensures that the law evolves with technological advancements while protecting the integrity of the First Amendment.

---