Nina To/The Cougar
In August 2026,the 7th U.S. Circuit Court of Appeals ruled that a federal law criminalizing the private, in-home possession of obscene, entirely virtual AI-generated child sexual abuse material, or CSAM was unconstitutional as applied to the defendant.
The court relied on multiple Supreme Court precedents concerning privacy, obscenity and the First Amendment.
This decision demonstrates how outdated legal precedents can leave courts struggling to address technology that did not exist when those decisions were made. It is unacceptable that U.S. law has not kept pace with AI’s ability to generate increasingly realistic abusive material.
The facts of the case
The decision came in United States v. Anderegg.
The facts of the case are as follows: Federal prosecutors accused a Wisconsin man of producing “hyper-realistic images of prepubescent children engaging in sexually explicit acts” using the generative AI Stable Diffusion.
Anderegg, a software engineer with a background in AI models, was charged with producing, distributing and possessing the material, as well as transferring it to a minor younger than 16.
Obscenity versus the right to privacy
A federal district court dismissed the possession charge and the 7th U.S. Circuit Court of Appeals upheld that decision due to several several Supreme Court precedents. First, the landmark 1969 case Stanley v. Georgia established First Amendment protection for the private possession of obscene material in the home.
However, the circumstances in Stanley were significantly different from those in United States v. Anderegg.
People have a legitimate expectation of privacy in their homes. However, virtual CSAM should fall within the limited circumstances in which the government can restrict private possession.
Researchers have raised concerns that engagement with virtual CSAM could contribute to escalation in some offenders, including progression toward material involving real children or contact offenses. However, the relationship remains an area of ongoing research.
Obscenity in relation to fictional CSAM
The second Supreme Court precedent Lee cited was the 2002 case Ashcroft v. Free Speech Coalition. This case struck down some restrictions of the Child Pornography Protection Act. Most notably, the Court ruled that certain computer-generated CSAM could not be prohibited on the same grounds as material involving real children.
Because virtual images do not require the abuse of real children during production, the government’s interest in protecting children from that specific harm did not apply in the same way.
Again, this outdated logic from 2002 does not work in a world of realistic generative AI. Although the children in AI-generated CSAM may not be exact copies of real minors, AI uses existing sources to create new content.
Therefore, every AI-generated depiction of CSAM is re-exploiting real victims, whose faces and bodies are scraped from a database and stitched together into a grotesque Frankenstein of obscenity. Moreover, giving any hyperrealistic CSAM a pass will make it harder to identify real cases of abuse.
If lifelike CSAM is alright as long as it is technically fictional, the line can be blurred quickly. If an AI image depicts a real child but with differently colored eyes, is that a completely different child?
Lack of government enforcement creates loopholes
The appellate opinion, written by Lee, expressed concern about the ruling. He acknowledged that the Supreme Court precedents behind the decision were established “nearly twenty-five years ago,” when today’s image-generation technology was “likely unimaginable.”
Lower courts are bound by Supreme Court precedent, and the 7th Circuit said it was not free to redraw the lines established by those earlier decisions despite its concerns about advances in AI.
Congress tried to address the problem further by creating the ENFORCE Act, which is designed to update the laws around CSAM and AI.
However, that act was introduced in 2025 and a year later, it still has not become law, despite being bipartisan. Meanwhile, the pace of technological development continues to outstrip the legislative process.
The First Amendment was never designed with generative AI in mind. Nearly four years after image-generation tools entered the mainstream, courts are applying decades-old precedents to technology capable of producing material the Supreme Court could not have envisioned. If the law cannot keep pace with that technology, the gaps will only become more difficult to address.
opinion@thedailycougar.com
