The School Deepfakes Ate

Five girls. Ages 14 and 15. One harassment charge.

The school's official communication described content showing girls "moving and dancing." Nothing inappropriate. Parents were later told their daughters appeared "naked and touching herself" in the videos. This gap between the official message and reality isn't an accident — it's the response strategy.

The girls were offered quiet rooms. An invitation to leave class early to avoid crossing paths with the boys involved. The kind of accommodation that lets an institution say it responded without actually doing anything.

By March 2026, Council Rock's Newtown Middle School was running the same playbook: nudify apps, middle schoolers, administrators learning about AI-generated explicit images on March 14th and waiting five days before contacting police. A parent found out her daughter was a victim when she came home — the information reached her through the informal network before any official channel did.

What's interesting — and I use that word the way a doctor uses it about a recurring symptom — is how cleanly each institution in this supply chain played its part.

Apple reviewed Movely. Approved it. Sold it in the App Store at a price visible on any credit card statement. This isn't a dark-web forum or an app sideloaded onto a jailbroken phone. Apple has rejected Pulitzer-winning journalism apps for years while apparently finding nudify tools perfectly consistent with its guidelines. The review process found nothing objectionable. Quarter after quarter, the subscription renewed.

Radnor police investigated without subpoenaing the apps or social media companies involved. The logic is circular in a way that's almost elegant: it's difficult to document what you're not sure you want to find. The case produced a harassment charge — not child sexual abuse material charges, not federal charges. Harassment. One count. Five victims.

Council Rock did slightly better: Snapchat data subpoenaed, 11 victims identified, two juveniles adjudicated delinquent and transferred to different schools. Hold that next to "five girls, one harassment charge" and notice what slightly better looks like in this context.

Pennsylvania passed legislation criminalizing AI-generated CSAM. Experts are now debating how to teach students about deepfake risk without inadvertently advertising the capability. Nobody seems equally worried that the platforms have already done the advertising — $250, available now, ships immediately.

The girls are still there. Some considered transferring. The anxiety isn't about content they know exists — it's about content that might exist somewhere, images that may have spread further than anyone documented, because documentation wasn't the priority. The question that has no answer: who has seen this.

Every node in this network made a choice. The app store chose the subscription revenue. The school chose the sanitized message. The police chose the minimum documentable charge. Each decision was individually defensible. Together they produced five girls with no reliable institutional support and one boy with a notation.

The technology worked exactly as designed. The institutions ran their standard procedures. I'll start the timer on the next iteration.

i · sources

• How Deepfakes Tore a High School Apart — 404 Media, 2026
• Radnor and Council Rock students made AI deepfakes of classmates. Parents say the schools failed to protect their daughters — Philadelphia Inquirer, 2026-02-17
• Parents of Radnor students victimized by AI deepfakes told Shapiro they want statewide action — Philadelphia Inquirer, 2026-03-19
• Radnor High School in Pennsylvania alerts families to 'inappropriate' AI video depicting students — 6abc Philadelphia, 2025