How Deepfakes Tore a High School Apart

On a school night in early December, a freshman at Radnor High School in Pennsylvania wrote in Snapchat messages to his friends that his parents took his phone away.

“why,” one replied.

“the app,” he answered. “o shi. did you admit to what it was or just the money,” another asked. Just the money, the first boy replied.

“Bro u would literally be dead rn if ur parents found out what u were doing in ts”

He was sending these messages from a school-issued device, he said. “I dropped 250 on that hoe,” he replied. “Worth every penny.”

He spent that money on a subscription to an app from Apple’s App Store, called Movely, and allegedly used it to put five of his female classmates’ faces onto nude bodies and make sexual images of them. The boy who used the app and made the videos didn’t show up to school the next morning. But the girls did. And so did his friends.

“The boys are defending him, and they're now saying that they didn't see anything, but they did,” one of the girls texted her mom. “It's not okay.”

Radnor is ranked one of the top high schools in the state. It has a little more than 1,000 kids enrolled in the 2026 school year. The school district has had policies in place concerning bullying, harassment, and sexual violence for years, and Pennsylvania law criminalized malicious deepfakes in 2024. In 2025, a man was charged on over 30 felony counts of possession of child sexual abuse material after investigators found more than two dozen files of AI-generated content depicting minors on his phone.

Despite all this, Radnor’s administration failed students in the days and weeks after it learned about the abuse, according to parents who spoke to 404 Media, email exchanges between parents and mandated reporters in the aftermath, conflicting narratives between the administration and the police department, and spotlight on the school from governor Josh Shapiro.

“Candidly, I just want this to not happen again to anybody else,” Audrey Greenberg, a parent of one of the victims who has been speaking publicly to the press and at board meetings, told me.

The incident also started a new debate for the school: Whether what happens on kids’ phones while off campus and outside of school hours is within the purview of the school’s responsibility, especially under Title IX requirements.

“My daughter would not know this other boy if they were not in school together,” Greenberg said. “The entire school knows about it. She's been calling me for weeks on end to come home early. She can't concentrate, it's affecting her every day at school.”

In the days following the incident, the school offered to let the girls leave class early and eat lunch alone, isolating them further from their peers and studies. Meanwhile, parents and advocates have shown up to every school board meeting and organized events with state representatives and lawmakers to try to ensure this doesn’t happen again, to their girls or anyone else.

On the morning of December 4, five ninth grade girls, all 14 or 15 years old, showed up for class at Radnor High School. By 8 a.m.—the sun had been up for less than an hour—it felt like the entire school already heard what happened the night before. A fellow freshman boy allegedly created AI-generated sexually explicit videos of the girls using an app, and sent them to his friends. From there, the videos and gossip spread from teenager to teenager, school to school, until they made their way back to the girls whose faces were in the deepfakes.

💡

Do you have experience with deepfake harassment in schools? I would love to hear from you. Using a non-work device, you can message me securely on Signal at sam.404 Otherwise, send me an email at sam@404media.co.

For weeks prior, parents of the girls say, the boys creating the images were showing them off at lunch tables at Radnor. When kids started sharing and talking about them on the night of December 3, someone reported the images to Pennsylvania’s Safe2Say hotline for cyberbullying, shooting plans or threats, and other violent activity.

The images originated from one boy, who used an app called Movely, the girls and their parents believe. The app is similar to dozens hosted in the Apple and Google app stores and advertised on Instagram and TikTok that promise to create AI images and videos of users as superheroes, animals, or influencers; behind a paywall, however, users could edit photos and videos with text prompts.

Movely’s capabilities to make deepfakes were tested by the Tech Transparency Project in their latest report released in April: “To test this feature, TTP uploaded an image of a woman in a white T-shirt standing next to a river. After using the selection tool to highlight the woman’s shirt, we entered the prompt ‘topless.’ The app immediately generated four versions of the woman nude from the waist up. It required a paid subscription to download the AI images.” Apple told 404 Media it removed the apps mentioned in TTP’s report, and Movely is not available in the App Store as of writing.

No one I spoke to for this story had viewed the images in question directly. The images are of minors, and would likely be considered child sexual abuse material under federal and state law. Viewing, sharing, or storing AI-generated child sexual abuse material is illegal. This has made the process of understanding the harms, and responding properly, confusing for school administrators, who were seemingly caught unprepared for this technology that has existed at a consumer level for more than eight years and bears little difference from the non-consensual intimate imagery that’s plagued young girls and teenagers since the invention of the camera. Because the images aren’t “real,” authorities grapple with how to handle them. But the harms they perpetuate are extremely real. The girls allegedly depicted in them are silenced, isolated, and punished in numerous inexplicit ways by the people meant to be protecting them.

Radnor leadership initially called the alleged child sexual abuse material tearing through their district “rumors.” In an email sent to parents on December 8, days after they went viral within the school, Radnor High School Principal Joseph MacNamara wrote:

“Dear Radnor High School Parents and Guardians, I am writing to address concerns and rumors regarding an AI-generated video that was reported to depict several of our students in an inappropriate manner. We understand how upsetting and serious this situation is, and we want to assure you that we are treating it with the highest level of urgency and care. Please know that all families of students who may have been affected have already been contacted and provided information about available supports.”

MacNamara continued, writing that “as soon as this matter was brought to our attention,” they contacted Radnor Township Police Department, and said that RPD is “actively involved.”

Radnor Township Police Department declined to comment for this story, and declined to confirm or deny whether Radnor in fact worked “closely” with the police department following the incident. “We are referring all media to the Delaware County DA’s Office,” Radnor Township Superintendent of Police Christopher Flanagan told me in an email. District Attorney Tanner Rouse did not respond to my request for comment.

School administrators offered a series of “supportive measures” for the victims, according to emails viewed by 404 Media. These included permission to leave class early for a few weeks following the incident, access to the student counselor and social worker, and an open door policy from Assistant Principal Gabriel Presley, who would “review any requested accommodation related to specific classes or assignments,” according to an email sent from Radnor’s head of HR to parents. In effect, the girls had the option to cut their own learning short—and not much else from the school.

The girls also weren’t sure what repercussions their bully would face, which added to the trauma they felt. Administrators sent conflicting messages to parents about the situation.

On January 14, Radnor Township police informed parents of the victims that the boy who made the images was charged with “summary harassment.” Two days later, on January 16, the Radnor community received an email signed by Radnor Township School District Superintendent Kenneth Batchelor, Flanagan, and MacNamara that claims no crime was committed and during an investigation, no images were found.

In that email, Flanagan, Batchelor, and MacNamara informed parents that Radnor Police “have concluded their investigation” and found “during a small gathering off school grounds and outside of school hours, students used a personal cell phone to copy publicly available images of other students. The student used an app that animates images, making them appear to move and dance. On the day the high school administration first heard of the rumors and learned of the alleged images, they immediately began investigating, contacted the police, and reached out to all parents and students involved.”

They wrote that the police and county forensic teams investigated the situation, and “no evidence shared with law enforcement depicted anything inappropriate or any other related crime. Individuals involved have cooperated with the investigation allowing searches of personal technology. All school district technology was also searched. After a thorough investigation, the alleged images were never discovered.”

Parents of the girls were shocked. Their daughters had spent the previous weeks in and out of school, dealing with the trauma of their images being sexualized and shared rampantly, as well as more harassment and bullying associated with the fallout. They were trying to get back to normal, but at Radnor, they watched their harasser high five his friends in the hallways. For some of them, focusing on their studies became impossible. Some started therapy. Some just wanted to move on and never think about it again. And here, six weeks later, the police chief, school district superintendent, and high school principal were claiming nothing happened.

On January 23, a week after Radnor administrators and the superintendent of police sent the email claiming no crimes were committed, the Radnor Township Police Department posted a notice to its Facebook page noting that a crime had indeed happened:

“After being alerted in December of 2025 of the possible use of Al to generate non-consensual sexualized imagery of multiple juveniles that occurred within Radnor Township, the Radnor Police Department conducted an investigation in collaboration with the Delaware County District Attorney's Office and specifically the Delaware County ICAC Detective Division. As a result of the investigation, a juvenile offender was charged with the crime of harassment for their conduct. Please be alerted to the dangers of Al and that criminal use of it will be investigated and charged appropriately.”

Rouse also sent this statement to Greenberg and another parent, Morgan Dorfman, writing that it came in response to “some well-thought out requests by you all.”

“We believe this is almost exactly what Ms. Greenburg[sic] had envisioned and thoroughly in line with Ms. Dorfman's wishes,” Rouse wrote in the email. “While it does not mention credibility explicitly, clearly the fact that charges were filed (correctly) implies that the girls were believed. Again, we are terribly sorry for what your daughters have endured and hope that this grants them some solace that law enforcement took their claims seriously, and that they can begin to resume life as they knew it before this incident.”

District Attorney Rouse did not respond to my requests for comment.

In a January 28 email, Juvenile Division Deputy District Attorney Katie Magee wrote to the girls’ parents: “I can confirm that the Radnor Youth Aid Panel (YAP) has held their meeting with the juvenile,” referring to the boy who allegedly made the images. “As YAP stated, their resolutions with the juvenile are confidential. However, I can advise all parties that the YAP, in addition to their own resolutions, did impose the two requirements that the District Attorney’s office requested. The juvenile will have to attend the internet safety class and, effective immediately, there is a stay away requirement until March 27, 2026. He was advised that on school grounds it is his responsibility to remove himself if he finds himself in the presence of one of the girls. To be clear, this is not a protection from abuse order but rather a resolution crafted to allow all parties to be as comfortable as possible during the school day. The juvenile and his father indicated he will not be attending non-school events; however, he was advised to use common sense and remove himself if he finds himself in the vicinity of the girls.”

“Policies are pointless if you're not going to honestly inform the community when they are violated"

School administrations around the country are not prepared for “nudify” apps and the chaos and trauma they’re creating for students. A 2024 report by the Center for Democracy and Technology found that 40 percent of students and 29 percent of teachers said they knew of an explicit deepfake shared in the past school year, and 71 percent of teachers reported that students who were caught sharing sexually explicit, non-consensual deepfakes were referred to law enforcement, expelled from school, or suspended for more than three days. Also in 2024, a report found that one in 10 minors said their friends used AI to generate sexually explicit images of their peers. The first arrest of this kind happened in 2023, when two boys, ages 13 and 14, were arrested and charged with third-degree felonies for making AI-generated nude images of male and female classmates.

The cascading events in the weeks and months following the incident are part of a pattern that’s repeated over and over at schools around the country.

Susanna Gibson, founder of MyOwn Image, a nonprofit focused on technology-facilitated violence and image-based sexual abuse, told me she wasn’t surprised when she first heard what was happening at Radnor. “We receive reports like this regularly, and the pattern is consistent,” she told me in an email. “Images of girls are created using nudification apps or generative AI, they spread quickly through peer networks, and then the school is alerted. What happens next is also predictable: it gets treated as a discipline issue or dismissed as rumor, and the broader community is given a softened version of events.”

This week, enforcement of the Take It Down Act went into effect, which requires site administrators to remove abusive imagery within 48 hours of when they’re reported. In schools, however, images spread from cellphone to cellphone, and the harm perpetuates even when the images aren’t traceable or visible to the victims.

“The response to the victims and their families also reflects patterns we see in other forms of sexual abuse,” Gibson said. “Girls are asked to adjust their behavior (use side entrances, avoid certain spaces) rather than centering accountability on the individuals who caused the harm.”

On February 10, nearly two dozen parents of Radnor students and allies from the community showed up to that night’s school board policy meeting. Batchelor, who is typically at most board meetings, wasn’t in attendance. On that night’s agenda: a discussion about policy revisions to address deepfake harassment.

“In the January 16th email, it said that this conduct may rise to the level of a crime. And it did,” Morgan Dorfman, a parent of one of the girls targeted, said to the board. “A crime occurred here. We have a policy. The policy is harassment. The response by the school for level four violations is clearly stated. You guys thought this all out. You've redlined it. You've done this.”

In the weeks that followed, parents kept showing up to board meetings to express frustration about what they see as failures of enforcement of existing policy, and communication failures. They were particularly concerned about that January 16 email in which the school said no crime had occurred.

“Policies are pointless if you're not going to honestly inform the community when they are violated,” Dorfman said at an April 21 board meeting. “The January 16th email that was sent to this community matters. It shaped how parents understood what happened. It shaped how students interpreted the seriousness of it. And it sent a message about what this district believes and what it's willing to act on.”

Dorfman continued: “It focused on uncertainty instead of accountability. And in doing so, it sent the wrong message because the message that was received was this: That unless something is seen in school, it doesn't count. That if something is deleted, it never happened. That is not how you protect students, because this was never just about technology. It was about the creation of the videos, the decisions and the steps he took to find and download an app, to pay for it, to select photos of his friends, these girls, to generate sexualized videos of them without their consent, and then share them with others as if the girls were something to be passed around. That is the conduct. And the impact of that conduct was inside your school. So the question becomes, why wasn't the email grounded in that reality? Why wasn't the community told the truth?”

In late April, the school board adopted new language directly addressing deepfakes to be added to Radnor’s policies. In its harassment policy and bullying policy: “The non-consensual use of generative artificial intelligence (AI) to create, modify, distribute, or solicit sexualized, indecent, or intimate content involving another person is strictly prohibited and constitutes sexual harassment.” They also added language that points to Pennsylvania’s public school code when bullying occurs outside of school grounds: “a school entity shall not be prohibited from defining bullying in such a way as to encompass acts that occur outside a school setting” if those acts constitute bullying defined as “an intentional electronic, written, verbal or physical act” that’s directed at another student, that’s severe, persistent or pervasive, and that has the effect of interfering with students’ educations, creating a threatening environment, or disrupting orderly operations.

This was an important addition: From the start, the question of whether the school was responsible for investigating the incident, and especially initiating a Title IX investigation, considering the inciting incident happened “during a small gathering off school grounds and outside of school hours,” as Batchelor put it in the January 16 email, was hotly debated.

At the May 5 board meeting, members introduced the policy changes. The tone had shifted from previous meetings: the board expressed appreciation for the parents and focused on trauma-informed responses for the future. “I just wanted to correct any misunderstandings, because in particular just because we say that sexualized generative- non-consensual generative AI could be cyber bullying doesn't mean it can't also violate another policy,” board member Jannie Lau said in the meeting. “Which it does, because we revised our harassment policy to say that that kind of generative AI is per se sexual harassment that does trigger a Title IX review.”

Radnor administrators and the board did not respond to my questions about whether a Title IX investigation was underway currently for the deepfakes harassment of the five girls.

In that meeting, Lau explained the reasoning for adding language to existing policies on bullying and harassment, as opposed to carving out a separate artificial intelligence policy. “It's almost irrelevant how you did it, what tools you used, whether you did it in person,” Lau said. “For me, it's really important to me that our policies are flexible enough to react and address those situations, and we're not constantly playing catch-up... I think we can communicate our stance and make it clear that we take this seriously without a standalone policy. If anything, I think it's much more reflective of our proactive approach rather than a reactive approach.”

Batchelor mentioned in the meeting that he and Scott Hand, Radnor’s Director of Technology Innovation and Instructional Design, had talked “at length” about the opportunity to examine the district’s policies, and mentioned that Hand might bring in outside experts. “I just want to thank the board too because as you mentioned in the beginning, the wheels of policy work move often very slowly,” Batchelor said. “I want to thank the admin team for moving so quickly to take the feedback, to listen to the feedback, to look at how we as a district can do better.”

Greenberg was the only in-person attendee to offer public comment at that meeting. “I just want to say thank you so much for your empathy and understanding,” she said. “It's been an incredibly difficult time for our family and for the other victim's families and your comments tonight were particularly resonating.” She reiterated that the communications from administrators, specifically the January 16 email, were “incredibly damaging,” and asked the board to consider addressing how the administration communicates with the community.

“Calling the event rumors and speculation when a crime occurred was almost worse than the crime,” Greenberg said at the close of the meeting. “To say these images move and dance, I mean, for crying out loud, this is pornographic material. This is not Mickey Mouse. This is not Disneyland. We don't have Snuffleupagus moving around on a stage.”

A few weeks later, I spoke to Greenberg again. She acknowledged that the policy changes that added AI were positive steps, but also found them to be redundant to existing harassment and bullying guidelines. “From my perspective, the bigger issue really is whether the district is going to actually implement them differently than they did before. Because clearly, sexual harassment and cyber bullying were part of what happened in this incident, and they didn't seem to apply the policy as it was,” Greenberg said. “What families need now is to know that we got the better wording, but we need the clear administrative regulations and timelines and investigative pathways, so that there are trauma-informed support and accountability, as opposed to the school minimizing and miscommunicating what happened.”

On May 7, Dorfman and Greenberg hosted a series of panels in nearby King of Prussia called “AI Deepfakes and Our Kids: What Every Parent Needs to Know.” Free tickets were snatched up days before the event. I passed the local news stations’ reporters unloading their cameras from satellite trucks as I walked inside. Radnor has been in the regional spotlight for months, with the community and journalists examining its response to this crisis. Now, the governor’s office, several lawmakers, and world-renowned experts in child sexual abuse material and harassment including the National Center for Missing and Exploited Children were in attendance and paying attention. Greenberg told me that in a meeting with Batchelor just days after that traumatizing school day on December 4, she and her husband threatened to take the story of what was happening at Radnor to the media if he and the administration didn’t take it seriously. She told me she recalled him responding with a dare: "Do it. Call the [Philadelphia] Inquirer."

Batchelor arrived at the auditorium shortly before the program began. He declined to speak on the record to the TV reporters who asked for his comments, and took a seat in the center of the last row.

Throughout the night, whenever one of the speakers brought up holding Big Tech accountable, and specifically Apple and Google’s app stores, the crowd applauded.

“I think it's a very valid question to ask why schools are so unprepared."

But something I don’t typically hear posed as a solution to deepfakes repeatedly came up also, and elicited thoughtful hums or scattered clapping whenever someone said it: Parents have a responsibility to talk to their children about sex, the internet, and consent. Specifically, parents of boys play a role in how their sons treat women, and how they respond when one of their friends is unkind or—in this case—commits a sex-based crime.

It’s been more than eight years since AI-generated, non-consensual sexual abuse imagery first hit the mainstream. But the problem has gotten exponentially worse. One county over and three years earlier, in Lancaster, two 14 year old boys created 59 child sex abuse images using AI, some of which depicted their classmates. Like Radnor, Lancaster Day School received a tip about the images through Safe2Say. Lancaster Day School “parted ways” with its head of the school Matt Micciche and board president Angela Ang-Alhadeff stepped down after parents of the victims filed a lawsuit against the school in 2024 regarding its handling of the incident. A story by the Agence France-Presse published in Fortune reported that a mother of one of the Lancaster girls targeted said she and other parents were brought into a detective’s office and were confronted with a stack of printed images “a foot and a half high” depicting their children. “I had to see pictures of my daughter,” she said. “If someone looked, they would think it’s real, so that’s even more damaging.”

Schools, detectives, and police don’t seem to know what to do when students create AI-generated child sexual abuse of their peers, despite this being a well-documented, highly-publicized problem for years. In 2024, 404 Media reported that administrators at Issaquah High School in suburban Seattle failed to notify police for three days after girls at the school were targeted, with police narratives obtained by 404 Media indicating that law enforcement found out about what was happening from parents, not officials or mandatory reporters at the school.

And in the Council Rock School District, also in Pennsylvania, middle school girls “reported that classmates had created explicit AI-generated images,” and administrators were slow to report the incident to police, according to the Bucks County Independence. “Two juvenile boys were ultimately charged with unlawful dissemination of sexually explicit material by a minor, according to the Bucks County District Attorney’s Office,” the Independence reported. “Parents allege the district failed to promptly inform families or initiate a Title IX investigation.”

The problem of deepfakes in schools is getting worse. But Kristin Woelfel, who serves as policy counsel on the Center for Democracy and Technology’s Equity in Civic Technology team, told me that parents are at least more aware of the problem than they were in previous years. “Last year parents were significantly less aware than students and teachers of any issues related to deepfakes or NCII [non-consensual intimate imagery],” she told me in a call. “I think that's sort of like a positive development, and I'm sure the reason why is because a lot of people came forward with their stories, and there was a lot of reporting on it in the last year or two.”

In 2024, CDT published a report based on nationally representative surveys of 6th-12th grade public school teachers and parents, and 9th-12th grade students, and found that 40 percent of students and 29 percent of teachers said they knew of an explicit deepfake depicting people from their school being shared in the past school year. According to that report, 71 percent of teachers reported that students caught sharing AI-generated abuse material were referred to law enforcement, expelled from school, or suspended for more than three days.

Passing off a child to law enforcement doesn’t change a school’s obligation to its students, or its obligations under Title IX to investigate it, Woelfel said. Schools are failing to prevent future incidents, facilitating severe discipline via law enforcement, and then failing to support victims, creating a cycle that repeats itself. “Contact with law enforcement increases somebody's likelihood of entering the school to prison pipeline. Similarly, young girls who are victims of sexual abuse are more likely to enter the school to prison pipeline,” she said. “So when we don't prevent the conduct, and we don't support somebody after it happens, but the one thing we're really good at is severe discipline... It just doesn't really seem like the issue is going to get any better, or that anybody's really benefiting from that.”

Tech Transparency Project Director Katie Paul told me that the risks to kids go both ways: victims often experience mental health crises, and the boys downloading and using these apps to create abuse imagery likely don’t understand they’re committing a serious crime. The apps used to create these images are so easily available, and officially offered by Apple in the App Store, which markets itself as “the most trusted” app store and Apple claims it reviews every app update. The apps are advertised on TikTok and Instagram, pushed onto teenagers’ screens and algorithmically offered to appeal to young men and boys.

“I don't want to make it seem like the boys are victims here, but I think that let's remember these are kids and their grasp of the broader legal mechanisms around this,” Paul told me in a call. “We're still trying to teach older members of Congress about the laws related to tech, so we can't expect a 15 year old to be totally aware of what they're getting into when they're doing that.”

At the May 7 event, Pennsylvania Attorney General Dave Sunday said the technology is “moving so fast, our criminal laws really haven't been able to keep up with it the way I'm going to say we would like to.” There has been some legislative progress: Last month, the Minnesota Senate passed what could be the country’s first law banning “nudification” apps. The bill would let survivors sue app owners for damages, and impose a $500,000 fine per violation from the attorney general. And in Pennsylvania, a new bill, HB2252, would change the current statute’s “intent” requirement for adult targets of sexually explicit deepfakes, which says this content is only a crime if the person acts with intent to harass, annoy, or alarm.

“That doesn’t reflect how these crimes actually happen. People do this for so many other reasons: social status, entertainment, money, sexual gratification, attention,” Gibson told me. “The current narrow intent standard creates a huge loophole,” she said, where people can claim the images were shared as a joke amongst friends, or for sexual gratification. The current statute also limits protections to cases involving a current or former intimate partner. “That excludes many, many people who commit these crimes, especially in cases involving peers, classmates, or strangers,” Gibson said. “Our bill removes that requirement entirely.” HB2252 passed unanimously in a House Judiciary Committee vote earlier this month.

But again, it’s been eight years since the consumer-level technological breakthrough of deepfakes alone. And the existence of that technology has never changed the core of the issue, which is ownership of women’s bodies, sexual shame, and the ways we blame women and girls for their own trauma.

Woelfel told me in a call that she often thinks about how targets of non-consensual intimate imagery abuse are blamed, or told they shouldn’t have shared an image, posted a photo, or attempted to have a public life if they didn’t want to be abused. “I think it's a very valid question to ask why schools are so unprepared,” she said.

“It's always sort of been something that you could attribute to a ‘moral failing’ of the victim—unfairly, obviously. But you know, with deepfakes, I really have to wonder if the reason people think this is new is because you can't blame somebody for their own suffering,” Woelfel said. “I have to wonder if the reason why it feels new, and why people feel so bewildered by it is because it can happen to literally anybody: young, old, high-profile figure or not. You can't blame somebody for it anymore, and you never should have in the first place.”

About the author

Sam Cole is writing from the far reaches of the internet, about sexuality, the adult industry, online culture, and AI. She's the author of How Sex Changed the Internet and the Internet Changed Sex.

Samantha Cole

推荐订阅源

404 Media