How AI has made child sexual abuse more prevalent

It all begins with a single screenshot from Instagram, Snapchat or some other social media.

The photo itself is nothing remarkable — a casual selfie, a holiday memory, a sports photoshoot, a prom photo. You know the kind. Billions of photos like these are shared on social media every single day. These are the kind of photos that get printed out and framed by grandmas all over the world.

But this photo — instead of getting a like, a comment or a special place on grandma’s wall — gets turned into something completely different.

For a 15-year-old sophomore in Illinois, it was a prom photo. Allegedly, two classmates at her school took a screenshot of her photo and used artificial intelligence tools to make synthetic nude images of her, dozens of other students and three teachers.

For another high school student, it was her yearbook photo. She was one of 60 girls at an elite private school in Pennsylvania who were targeted by two fellow students using AI software to create explicit photos and videos based on real images of them.

‘Nudification’

These are the stories we know about. They are not isolated incidents. Indicator reports there have been 116 reported criminal cases of AI “nudification” tools being used by students within school communities. The real impact may be much greater.

Recent research by UNICEF shows at least 1.2 million children reported having images of themselves manipulated into sexually explicit AI-generated deepfakes in the past year alone.

While deepfakes and image-altered exploitation have a long history, generative AI has made such deception and exploitation easily accessible for an increasing number of people, including teenagers. While nudification apps predate Elon Musk’s Grok chatbot, it was Musk’s intentionally edgy AI that first showcased AI’s sexually abusive abilities to the public.

“AI has made such deception and exploitation easily accessible for an increasing number of people, including teenagers.”

I first wrote about this for BNG in January, noting at the time it was a predictable outcome from a generative overseen by a known misogynist whose companies have a culture of sexual harassment. One analysis of Grok’s output over a 24-hour period found users had Grok generate about 6,700 sexually suggestive or “nudified” images every hour.

Of particular concern is that some users were asking Grok to sexualize or nudify images of children.

One of the earliest victims of Grok’s lack of guardrails was 14-year-old Stranger Things star Nell Fisher. Just a few weeks later, the national police of Ireland reported they were investigating 200 instances of child sexual abuse-related images generated by Musk’s AI software. In response, X restricted Grok’s image generation capabilities to paid subscribers, begging the question as to whether they were preventing misuse or monetizing it.

X was not alone in its tacit promotion of AI’s “nudification” capabilities. Meta, while drawing guardrails around its own AI image generation service, nonetheless allowed at least 11,000 advertisements for nudify apps to run on its Facebook and Instagram platforms.

Researchers repeatedly have found these nudification apps aren’t hanging out in sketchy and inaccessible corners of the dark web but are easily available in the regular Apple and Google App stores. This ease of access, coupled with the increasing ubiquity of artificial intelligence use, has led to the explosion of AI-generated child sexual abuse imagery created by children themselves.

Generative AI makes abuse easy

Peer exploitation and the sharing of inappropriate images is not new. A 2025 survey found one in 11 minors had created and shared a nude photo or video with a peer or social media follower. Of that number, nearly half cited being pressured into doing so.

A different study from 2023 found 21.6% of respondents age 9 to 17 reported having received “nude or near-nude” images of peers while 9.5% said they had an image of themselves nonconsensually reshared.

“21.6% of respondents age 9 to 17 reported having received ‘nude or near-nude’ images of peers.”

These are alarming statistics that should cause us to deeply and sincerely question the harm the ever-present ability to send and share images and video is doing to young people. But that number explodes when extended to AI-generated imagery.

A 2026 survey by digital forensics and clinical psychology expert Chad Steel sought to focus on the prevalence of creating and sharing sexualized images using generative AI. Drawing from a nationally representative dataset of American teenagers, Steel found more than half the study’s participants had created (55.3%) or received (54.4%) images generated using so-called “nudification” tools. Further, over one-third of participants reported having a nonconsensual image of themselves created (36.3%) or shared (33.2%) by their peers.

While we can’t directly compare these different studies of different U.S. teenagers at different times, it demonstrates a stunning escalation. Generative AI has not only introduced a new form of sexual abuse, but also dramatically expanded the scale on which abuse can occur.

With real imagery, roughly one in 10 young people (9.5%) reported having a nude or near-nude image of themselves nonconsensually shared. With AI-altered imagery, that number jumped to one in three (33.2%). Similarly, with real imagery, 21.6% of respondents said they have received sexualized imagery. That number jumps to 54.4% with AI-altered imagery.

Taken together, these figures point to a fundamental change in the scale and accessibility of child sexual abuse imagery. Generative AI removes many of the logistical barriers that previously limited the production of such imagery. Now there is no need to coerce or groom a victim. All you need is a single, regular photo of them. The result is a form of abuse in which nonconsensual creation and distribution can occur far more frequently and involve far more young people.

Courts struggle to keep up

It’s bad enough for classmates to misuse technology in this way. But the technology is available to adults with much more nefarious purposes. And the U.S. legal system is struggling to keep up.

A federal appeals court recently upheld a lower court’s dismissal of a child sexual abuse material possession charge against Steven Anderegg, a Wisconsin resident who allegedly used the generative artificial intelligence software Stable Diffusion to create photo-realistic AI-generated imagery of prepubescent children engaging in sexually explicit acts.

In October 2023, Instagram reported Steven Anderegg to the National Center of Missing and Exploited Children after he direct messaged two AI-generated images of nude and sexualized apparently underage boys to an underage minor in an apparent attempt at grooming them.

A warrant was issued for all Instagram accounts run by Anderegg, where investigators found he had sent a third image and engaged in sexually explicit messages with minors. A second warrant was issued for Anderegg’s home and electronic devices, where investigators found more than 13,000 AI-generated images, many of them appearing to be children engaged in sexually explicit conduct.

As a result, Anderegg was charged with four different CSAM-related offenses.

“They argued that private possession of the material is constitutionally protected by the First Amendment.”

Anderegg’s defense argued for all charges to be dismissed in district court. That court granted the motion on possession but denied dismissing all others. Anderegg could be charged with producing and distributing CSAM, but he could not be charged with possessing it.

Given that possession of something is a byproduct of producing it and a necessity for distributing it, this, at first glance, is nonsensical. However, Anderegg’s defense didn’t argue that Anderegg wasn’t in possession of the material, they argued that private possession of the material is constitutionally protected by the First Amendment.

The court agreed.

The outrage was immediate and bipartisan.

The left-wing media organization The Young Turks published a video clip of one of their programs on social media and titled it “Judge Says It’s OK To Victimize Kids Using AI, Sparks MAJOR Outrage.”

The right-wing religious group Focus on the Family published an article headlining that, because of an old Supreme Court ruling, virtual child porn was protected by the First Amendment.”

For at least this issue, political partisanship cedes to common sense. We understand inherently that this is wrong and that it harms children. Unfortunately, appeals courts must base their decisions not on common sense, but on Supreme Court precedent.

AI without guardrails

Existing prohibitions on the production and possession of child sexual abuse materials are based on legal argumentation that the generation of such imagery necessarily involves the exploitation of particular real-life children. Anderegg’s claim is that his AI-generated imagery does not involve such exploitation.

In 1996, Congress passed the Child Pornography Prevention Act. One of the particular measures of this act was to explicitly prohibit computer-generated images. In response, the Free Speech Coalition — a nonprofit trade association of the U.S. sex industry — filed a lawsuit stating the vagueness of certain phrases in the law were overbroad and could have an effect on the legal adult pornography industry.

Their focus was on the prohibition on creating images showing what “appears to be” underage children.

“What happens when a fictional image is so realistic that it is virtually indistinguishable from a photograph of a real child?”

In 2002, the Supreme Court considered Ashcroft v. Free Speech Coalition, where, in a 6-3 decision, they determined certain aspects of the Child Pornography Prevention Act are unconstitutional. One part of the ruling says, “The CPPA prohibits speech that records no crime and creates no victims by its production. Virtual child pornography is not ‘intrinsically related’ to the sexual abuse of children.”

In a very uncomfortable passage, it almost seems to suggest virtual CSAM as a solution to eliminating child exploitation, writing that “few pornographers would risk prosecution for abusing real children if fictional, computerized images would suffice.”

That ruling created the difficult question now presented by generative artificial intelligence. What happens when a fictional image is so realistic that it is virtually indistinguishable from a photograph of a real child?

In United States v. Anderegg, the government argued that modern AI changes the technological and law-enforcement landscape enough to justify treating these images differently. The Seventh Circuit all but affirmed the morality of the government’s argument while denying its legality based on the Ashcroft precedent.

Where we go from here

It is up to the Supreme Court and Congress to intervene. We have reached an era in which AI-generated imagery is indistinguishable from real life imagery. AI-generated CSAM may not be created from the harm of a child, but it does harm children.

Lawmakers must be willing to challenge the Supreme Court precedents tailored to an age that knew nothing of generative AI.

This extends from criminal pornographers to teenagers who think they are engaging in innocent fun.

Sociologist Danielle Hoyt writes that children and young people “often describe nudified images as jokes, curiosity or experimentation.” In other words, they treat nudification as a trend not unlike other viral image generation trends. They are not thinking critically or morally about the images they are producing. Their digital device serves as a moral barrier that distances themselves from the real harm that is produced.

The fact that an image is artificial can make the abuse seem less serious. Perpetrators are able to distance themselves from the real harm by justifying the creation or sharing of an image on the thinking that, while based in reality, it isn’t real. Further, AI has made image generation so accessible and easy that they do not feel any personal investment or connection to the image.

This allows the perpetrator to minimize the victim’s experience and their own responsibility. The social context can further reinforce this disengagement when peers treat the images as funny or trivial rather than recognizing them as sexual abuse.

“The bottom line is that nudification apps are wrong.”

The bottom line is that nudification apps are wrong. And while teens who use that technology to nonconsensually create and share images of their peers should be held accountable, the people, corporations and cultures that make the tools accessible to teens also should be held accountable.

Technology did not invent this impulse to control, humiliate, sexualize or objectify another person’s body. But it has made it easy, accessible and normative. Generative AI has removed the barriers that once stood between those impulses and their expression.

That means our response cannot simply be to teach children how to use AI responsibly. We must teach them how to treat other people responsibly. We must impress upon our kids — our sons, in particular — the real and substantive harm that deepfake nudes can do.

Tech companies continue to ignore or defend nudification. In August, the state of Minnesota enacted a first-of-its-kind ban on nudification apps. xAI, the creators of Grok, immediately filed a lawsuit to stop it. And yes, despite scrutiny, Grok is still able to create sexualized deepfakes. Google and Apple still host apps that are capable of nudification. We must cut off these inappropriate apps at their source.

We cannot accept a digital world where anyone’s body can be stripped of their agency with a few clicks, where companies profit from tools designed to facilitate exploitation, or where governments treat the harms of AI as a toy to play with.

Parents and educators have a responsibility to form young people who understand that another person’s body is not theirs to manipulate, sexualize or humiliate. Technology companies have a responsibility to stop building and distributing tools that make that violation effortless. And governments have a responsibility to establish meaningful boundaries when the market refuses to do so.

The goal is not to stop technological progress. It is to ensure progress does not come at the expense of human dignity.

 

Josh Olds is a public theologian and pastor for those disillusioned with institutional church. He is the creator of the small-group video series “Year on the Mountaintop” and a featured contributor to Fostering Hope: A Prayerbook for Fostering and Adoptive Parents. Follow his work on Facebook or at JoshOlds.com.