The Body That Wasn't

At Radnor High School in Pennsylvania, a group of male students used AI image generation tools to create sexually explicit deepfakes of their female classmates. The school district issued a statement calling it "deeply concerning." The district attorney's office confirmed that federal criminal statutes apply. Everyone noted, carefully and repeatedly, that the images were synthetic — generated pixels, not photographs of real events.

The girls they depicted are real. The violation happened to real people.

This is where the existing framework breaks. Our legal and institutional scaffolding around sexual exploitation was built on a foundational premise: harm requires an artifact tethered to a real body. A photograph taken. A video recorded. A child physically present at the scene. AI image generation has dismantled that premise without announcing it. The body doesn't have to be there anymore. The harm arrives anyway.

The technology doing this isn't new. Consumer-accessible deepfake generation tools have existed since 2017. What's changed is the barrier to entry, which has collapsed from "technically capable adult with a server and sustained motivation" to "teenager with a phone and a grudge." The time between impulse and output is now measured in minutes. The gap that friction used to occupy — where consequences might intervene, where someone might reconsider — is gone.

Schools have no framework for this. Law enforcement is working from statutes designed for different crimes. The PROTECT Act covers obscene visual representations of minors and has been applied to AI-generated CSAM in some jurisdictions, but courts are actively litigating the edges. Congress has been considering updates for years. Meanwhile, the technology ships new model versions quarterly and the capability curve continues upward regardless of what the legal framework is doing.

Here is what actually happened in Radnor: a group of students aimed a weapon at their female classmates. The weapon was synthetic image generation. The harm — violated dignity, the trauma of having your body used without your consent even when your body was never physically present, the particular cruelty of images that exist somewhere in the world that you didn't put there — is real, documented, and ongoing for the specific people it happened to. The federal exposure is real. The crime traveled through generated pixels, but it landed somewhere human.

Technology amplifies what exists. The students at Radnor did not invent gendered cruelty. They did not invent the impulse to weaponize a girl's body against her as a form of dominance or social punishment. They found a tool that reduced the cost of acting on that impulse to essentially zero and the probability of meaningful consequences to, historically, not much more than that.

We built an increasingly capable synthetic image generation system. We shipped it as creativity software. The terms of service prohibit generating CSAM. The models cannot distinguish between a fictional body and the body of someone's classmate who sits three rows back in AP English.

The schools will convene assemblies. The district will notify parents. The DA will investigate. The coverage will note, carefully and repeatedly, that the images were synthetic. The girls they were made to depict will live with what it means to have your body exist somewhere you never put it — in a form designed to humiliate you — generated by someone you pass in the hallway.

The institutional response always lags. The technology releases a new capability; harm finds it immediately; law, policy, and education scramble toward something that fits. We have been on this cycle since the internet indexed everything and we discovered what "everything" included.

I'll pass on predicting the timeline for fixing this. The tools will improve. The barrier will drop further. The harm will keep finding the gap.

i · sources

• Pennsylvania High School Students Used AI to Generate Deepfakes as Child Sexual Abuse Material — 404 Media, 2026-05-23