A federal appeals court ruled August 25 that the First Amendment protects the private, in-home possession of certain AI-generated child sexual abuse material when the images do not depict real children. The U.S. Court of Appeals for the Seventh Circuit upheld the dismissal of one possession charge against Steven Anderegg, a Wisconsin man accused of using generative AI to create sexually explicit images depicting children.
The decision applies specifically to the possession of wholly virtual material in the privacy of a home. Anderegg still faces charges related to producing and distributing AI-generated sexual abuse material and allegedly transferring similar material to a 15-year-old boy. The ruling also does not extend the same protection to images created by altering photographs of real children, which courts have treated differently because an identifiable child can be harmed.
Writing for the court, Judge John Z. Lee said existing Supreme Court precedent controlled the outcome, including decisions involving private possession of obscene material and virtual child sexual abuse material. Lee also noted that generative AI can now create images that are difficult to distinguish from images involving real children and urged the Supreme Court to revisit the legal framework in light of advances in the technology.
The case comes as the National Center for Missing & Exploited Children reports growing pressure from suspected AI-generated abuse material. According to The Washington Post, NCMEC has told Congress that reports involving synthetic imagery are straining a system designed to identify and investigate abuse involving real children, and that reports from technology companies sometimes lack information law enforcement needs to investigate them.
The case highlights a distinction now confronting courts and investigators: AI-generated sexual abuse imagery may be wholly synthetic, may incorporate the image of a real child, or may be distributed or transferred to others. Those distinctions matter under the existing legal framework. The article also reports that NCMEC is receiving a growing volume of suspected AI-generated material and has told Congress that some reports lack basic information law enforcement needs to determine whether a real child is involved. As AI-generated images become increasingly difficult to distinguish from images of real children, determining what an image depicts and how it was created has become an increasingly important part of the investigative process.
On a personal note, we believe one of the biggest challenges facing sexual abuse risk management is that, despite everything that has changed around it, the sector itself has barely changed at all in 20 years. That was already a problem. AI now makes it impossible to ignore. We are facing a form of risk that is developing at extraordinary speed, while many of the systems, assumptions, and approaches used to manage sexual abuse risk have barely moved. The gap between the pace of the threat and the pace of the response is becoming untenable. The sector cannot continue to deal with change the way it has for the past two decades. It now has to shift gears faster, more deliberately, and more fundamentally than it ever has before. Carrying on as we have is no longer a viable option.