When the victim isn’t real, is it legal to possess child sexual abuse images?

Artificial intelligence is forcing America's courts to confront unprecedented questions

What happens when technology can create an image that looks real, but no person was involved

And when those images depict child sexual abuse, are they protected by the First Amendment

42-year-old Stephen Anderegg was arrested in 2024. Prosecutors say he used AI to create thousands of photorealistic abuse images

Straight Arrow spoke with experts to understand the court ruling, limits to the First Amendment, and how emerging technology now complicates decades-old court decisions

On August 25th, the 7th U.S. Circuit Court of Appeals in Chicago

ruled that the government could not prosecute Wisconsin man Stephen Anderegg in a case involving entirely AI-generated images depicting fictional minors

But the ruling is much narrower than many headlines suggest. The court did not say that all AI-generated sexual material is legal

Instead, it addressed a specific constitutional question involving private possession of the images inside a private home

The Seventh Circuit affirmed the district court's ruling, where the district court had held that the First Amendment protects the purely private possession

of obscene material inside the home, even when that material is a depiction of imaginary children

Rihanna Thefferkorn is a policy fellow at Stanford's Institute for Human-Centered Artificial Intelligence, where she studies AI, law, and child safety issues. Anderig was charged

with producing, distributing, and possessing images depicting child sexual abuse. Images all generated by AI Federal prosecutors alleged that Anderad used Stable Diffusion a generative AI program to create the images The government also alleged that some of the

material was transmitted to a miner. Anderad faced several federal charges involving AI-generated

images, but this appeal focused on the possession charge solely. The Seventh Circuit did not rule

that creating or distributing the material is protected by the First Amendment

Instead, the judges considered a narrower question. Can the government prosecute someone for possessing entirely virtual material inside their home

The big distinction here, I think, is that even when it comes to realistic-looking depictions of virtual children

the Supreme Court has repeatedly said the distinction in terms of whether something is constitutionally protected or not

is was there a real child involved in the production of that material or not

And since there is no real child involved here, ultimately it becomes much harder to try and argue against

why that speech ought to be protected. And to answer that, the court had to look back decades

long before the development of artificial intelligence. In 1969, the Supreme Court ruled in Stanley v. Georgia

that the government generally could not criminalize the private possession of obscene material inside someone's home. But in 1990, the Supreme Court

created an important exception. In Osborne v. Ohio, the court ruled that the government could

prohibit possession of child sexual abuse material involving actual children. The justices said the

government has a compelling interest in protecting children from exploitation. Then came a critical

case in 2002. In Ashcroft v. Free Speech Coalition, the Supreme Court rejected efforts to broadly

treat fully virtual depictions the same as material involving actual victims That distinction became central to the Anderrig case The court didn even seem happy about having to say hey we have binding Supreme Court precedent on this question

and we don't like this material either, but we have to follow the Supreme Court on this issue

Are the judges themselves, you know, suspect here? And the answer is no. What makes this case so challenging is that today's technology looks dramatically different

from the technology that existed when those Supreme Court cases were decided

Kate Ruan directs the Free Expression Project at the Center for Democracy and Technology

and focuses on First Amendment issues involving emerging technology. Generative AI has changed the landscape by just making it a lot easier

to create child sexual abuse material that doesn't depict a real child

The advances in generative AI have done is it makes it harder for law enforcement

who are receiving reports of child sexual abuse material to identify children who are in harm and in need of immediate assistance

In Anderegg's case, the government argued that even entirely virtual material can create real-world harm

Among those concerns are whether it could be used for grooming, whether it normalizes abusive behavior

and whether increasingly realistic AI images make investigations more difficult. So I come to this work as somebody who has lived sexual abuse

has lived child sexual abuse, and also who works very closely with survivors

Lennon Torres works with the HEED Initiative, a nonprofit that's focused on protecting children

and survivors from online exploitation. Torres says that's why the human cost of this debate

is never far from mind. What it really does is normalize this issue. And so when you have

deep fake imagery it basically just rocket launches the amount of volume of things that are available to use to continue harming kids So what does the ruling actually do It protects Ander Egg from prosecution under the specific possession charge challenged in this case

But it does not make producing or distributing the material broadly legal

And it does not apply to material depicting an actual child. The law is based on decades old Supreme Court precedent

but generative AI has changed the technology landscape dramatically. That creates a gap between existing law and what AI can produce

The question now is whether the Supreme Court should revisit the rules. This type of technology essentially puts every child at risk

When we think about posting our kids on social media or we think about our kids out, you know, at the park

This technology has made it especially available for perpetrators to take just a normal, non-exploitative image of a young person and essentially use that to make really egregious and terrible content that they not only will consume, but often share

Torres says it's important to bring the law in line with today's technology

To put it bluntly, what people don't get right now when we're talking about this issue, what perspective is missing are the actual survivors of the imagery that these models were trained off of

And the reason that technology is able to do that is because it has seen and used the imagery of real children truly and horribly being raped in real life

And I think that we miss that part of the story and we forget that there are real people behind this

This case shows that while artificial intelligence is moving fast, the law is not moving at the same pace

Whether the law can keep up may be one of the defining legal battles of the AI era