Koffels News

A federal appeals court in the United States has recently affirmed an unsettling decision regarding AI-generated child abuse material.

The case concerns defendant/appellee Steven Anderegg, who was charged with possessing, producing and distributing visual material of minors engaged in sexual conduct, and with transferring material to a minor under the age of sixteen.

Anderegg’s activity came to light through a CyberTip Meta sent to the National Center for Missing & Exploited Children, following messages Anderegg allegedly sent via Instagram direct message.

He produced the material using Stable Diffusion, a generative AI tool, creating what the court described as hyper-realistic images of prepubescent children engaging in sexually explicit acts.

The judgment, decided on 25 August 2026, affirmed that in the United States, mere possession of AI-generated child abuse material that does not depict real children is not an offence in one’s own home.

The decision

Anderegg argued that being convicted for possession of child sexual abuse material would violate his First Amendment right. The First Amendment protects the right to freedom of expression and religion. It does not, however, protect material that is obscene.

Anderegg’s argument was based on the decision in Stanley v Georgia, 394 U.S. 557 (1969). In Stanley, law enforcement discovered reels of film containing obscene adult material at the defendant’s home. The Supreme Court held that the defendant’s First Amendment rights were violated, deciding that the state’s power to regulate obscenity “does not extend to mere possession by the individual in the privacy of his own home”.

In 1996, an Act was passed that extended the definition of child sexual abuse material to include any visual depiction, including those that are computer-generated, criminalising virtual child sexual abuse material.

However, Ashcroft v Free Speech Coalition, 535 U.S. 234 (2002), questioned the Act and found the provision violated the First Amendment. Importantly, it also rejected many of the arguments raised to justify the government’s prohibition on virtual CSAM.

In Anderegg, the court asked whether virtual CSAM was akin to real CSAM or, alternatively, obscene material. Given that arguments treating virtual CSAM as akin to real CSAM had already been refuted in Free Speech Coalition, the court ruled that virtual CSAM, for the purposes of this case, could be treated as obscene material.

The court therefore concluded that the AI-generated images fell within the category of obscene material, which meant the protections in Stanley applied. Anderegg was afforded the protection of merely possessing the material within his own home.

The judgment itself acknowledges concerns about the implications of this outcome, particularly in light of advances in AI technology.

Synthetic material, real harm

Legal commentary within the judgment, and following the decision, highlights aspects of AI-generated abuse material that the precedent in Stanley and Free Speech Coalition could not address.

Foremost, Free Speech Coalition, which refutes arguments on the likeness of real and virtual CSAM, reflected the technology of its time, namely the early 2000s. It does not substantially address how advances in technology have possibly increased the similarities between virtual and real material.

Judge Lee and Judge Kolar state that with present and evolving technologies, “AI-generated CSAM is virtually indistinguishable from CSAM created through the physical abuse of a child.”

A further significant issue is that many examples of real child sexual abuse material have been found within the datasets used to train AI. The dataset used to train Stable Diffusion, the software Anderegg used, was found to have hundreds of instances of child sexual abuse material.

AI-generated abuse material may not depict real children, but it is not victimless. It can rely on, and contribute to, the serious and widespread issue of child sexual abuse. We have written before about why AI-generated material is likely to increase, and about the parallel issues raised by AI tools built into mainstream platforms such as Grok on X.

Judge Lee and Judge Kolar expressly noted that additional guidance from the Supreme Court on the intersection between the First Amendment and virtual CSAM would be of benefit if an appropriate case were to arise.

What about Australia?

The legal landscape in Australia differs because Australia does not have a constitutional Bill of Rights and operates under different statutes and common law.

Presently, the Crimes Act 1900 (NSW) outlines child abuse material laws within New South Wales. A child for the purposes of “child abuse material” is a person who appears, or is implied, to be under the age of 18. Even if the material is AI-generated, creating, distributing or possessing material that depicts the abuse of someone under 18 constitutes a criminal offence. This is similarly criminalised under the Criminal Code Act 1995 (Cth).

Nevertheless, steps are being taken at a federal level to more explicitly and robustly prevent AI-generated child abuse.

Federal Parliament is considering a bill that would amend the Criminal Code Act. The Criminal Code Amendment (Using Technology To Generate Child Abuse Material) Bill 2025 aims to codify the issue nationally. The amendments would criminalise the use of a carriage service to download, access, supply or offer access to technology whose sole or dominant purpose is the creation of child abuse material, or to collect, scrape or distribute data for the purpose of training or developing such technology.

The Bill acknowledges that the measure will interfere with privacy, including personal property and the contents of a family home. However, it takes the position that this is lawful because the measure is not arbitrary, and is reasonable, necessary and proportionate to protecting the rights and freedoms of children, public order and public morals.

While the Bill is yet to be implemented, it would be a step towards stronger safeguards against AI-generated child abuse material. We have covered the broader shift in this area in our review of key 2025 reforms in child abuse law.

Why this matters beyond one case

The Australian Child Maltreatment Study (ACMS), published in 2023, demonstrates the striking prevalence of child sexual abuse. The study found over 1 in 4 Australians, 28.5%, experienced child sexual abuse before the age of 18.

The increasing incidence of AI-generated child sexual abuse contributes to an existing and far-reaching harm. Some research suggests