In every presentation I give to audiences of students and educators, I play two video clips side by side. Both show influencers endorsing products. One is real. The other is a deepfake. I ask the audience to guess which one is the fake. When I run this exercise with students as young as fourth grade, roughly 90% get it right. They spot the fake easily. When I run the same exercise with rooms full of adults, the majority often gets it wrong.
That surprises people, but the explanation is simple: Kids are swimming in AI-generated content every day. They’ve developed an intuition for it. They already know what deepfakes are. Many know how to make one.
In other words, the work of telling kids that deepfakes exist has already been done for us. What’s left for schools to do for students is twofold. The first part is deterrence: Students need to understand the real harm a deepfake causes and the serious consequences of creating or sharing one. The second is support: Every student should know that if they or someone they care about is targeted, they will be believed and they will be helped. Schools also need to ensure that the adults in their communities understand the scope and magnitude of the deepfake risk and are well-prepared to support students.
My consulting work with independent schools to prepare for and respond to deepfake abuse draws on my experience as a teacher, administrator, and technology ethics fellow at Stanford’s Human-Centered AI Institute. I believe it’s critical for independent schools to deliberately take on the deepfake conversation and train every layer of their communities from the board to the students. To help schools do this intentionally, I developed a deepfake safety course for students and lead sessions for school leaders, trustees, faculty and staff, and parents designed to inform, educate, and answer questions about their role in protecting young people.
The Risks Schools Face
When most people hear “deepfake,” they picture a doctored video of a politician or celebrity. But the deepfake risks most likely to reach a school look very different.
For example, every photo and video a school posts online is potential raw material for AI tools. Images of students function as biometric data that AI systems can use to identify, locate, or abuse them.
Schools also need to guard against AI-powered scams. Voice cloning, another type of deepfake, can replicate a person’s voice from a few seconds of audio. A cloned head of school’s voice could call the business office and request a wire transfer, for example.
The most urgent threat to schools, however, is deepfake sexual abuse. The harm to a victim is more severe than any other deepfake risk a school is likely to face. The law has also moved quickly: students who are targeted today have legal rights, including the right to have the images taken down, that did not exist two years ago. And right now, students are not hearing a strong message of deterrence from their schools. They need to understand that creating a sexually explicit deepfake of a classmate or a teacher is not only a violation of school rules but potentially a violation of the law.
In school communities, this most often begins close to home. A student pulls a classmate’s photo from social media or the school’s own website and runs it through a “nudify” app. These apps, which generate synthetic nude images from a clothed photo, are free, easy to find, and often marketed as pranks. The images often spread through group chats before any adult knows about them. They can be used to threaten or coerce students, and the resulting humiliation, fear, and violation of privacy can disrupt their academic, social, and emotional well-being.
Taking a Layered Approach to Education
Many schools’ first instinct is to add handbook policies outlining consequences for students who create deepfakes—and leave it at that. But deepfake incidents involve evidence handling, image takedown, victim notification, crisis communications, and insurance questions unlike anything most schools have faced.
Schools approaching this work should train each layer of the community in a deliberate sequence. The work should start with the senior leadership team first and should focus on establishing policies and response plans before the broader community is engaged. This is what McDonogh School (MD) leaders did as part of my work with them. “We wanted those closest to decision-making to sit with the complexity first,” says Robyn Little, senior director of educational digital strategy. “Before we could guide others, we needed to deepen our own understanding, engage our legal team, and assess our student reporting system.”
Board education should follow next and should focus on ensuring that trustees understand the school’s fiduciary and reputational exposure; using a tabletop crisis simulation to stress-test the response plan before an actual incident demands its use is a helpful exercise.
Faculty and staff should be next. They need practical guidance on responding when a student discloses an incident and should be prepared to reassure the student that they have done nothing wrong, affirm that coming forward was the right decision, explain that systems are available to help remove the content, and make clear that the school will support them. This preparation requires more than a slide deck and a handout. The training that McDonogh provided enabled teachers to process the information together and to workshop how they would engage students on the topic, Little says. That kind of active practice turns awareness into confidence.
Parent education should happen next. They will need a different kind of preparation that is focused on the concrete steps they can take immediately to reduce their family’s exposure, along with guidance on building the trust that makes a child more likely to turn to them before—or after—an incident. That trust matters because shame often determines whether a young person speaks up at all. A 2025 report from Thorn, “Deepfake Nudes & Young People,” found that 62% of young people said they would tell a parent if victimized by a deepfake. Only 34% of actual victims did. Two out of three kids carry it alone. Shame is what keeps them silent, and silence is what perpetrators count on.
Students come last in the education process––with different information at different ages––so that when they receive the message, they are surrounded by adults who understand the plan and their roles in carrying it out.
Talking to Students at Different Ages
Students need messages tailored to their developmental stage. The student sessions I lead address two age groups—grades four to six and grades seven to 12—with the content and framing adjusted for each.
Even fourth graders encounter voice-cloning deepfakes on platforms such as Roblox and Discord, AI-generated videos of trusted figures on YouTube, and AI companion apps that collect personal data. But they cannot set healthy boundaries. For them, the conversation is not premature; it is overdue.
For grades four to six, the session connects deepfakes to a framework that students already understand: the honor code. Creating or sharing a deepfake is framed as a form of lying and stealing, which helps younger students grasp its seriousness without legal language or case studies that are beyond their developmental readiness. They also learn what to do if they or someone they know is targeted and how AI companions that cannot set healthy limits may affect their real-world relationships.
Upper school students are ready for greater complexity: real cases; the legal landscape, including the TAKE IT DOWN Act (a 2025 law that criminalizes publishing nonconsensual intimate imagery); sextortion tactics and warning signs; and the privacy and emotional-dependency risks of AI companionship. They should also understand that using deepfakes to misrepresent themselves, such as by fabricating athletics highlight reels for college recruiting, can carry serious consequences.
“This is akin to taking the steps toward a more proactive approach to cybersecurity 10 years ago,” says Vinnie Vrotny, director of innovation and technology at The Kinkaid School (TX), another school I worked with recently. “It is the next potential threat exposure for our campus communities, and we owe it to our students and our faculty and staff to have put the work in and thoughtfully prepared.” Deepfakes, he says, are “a ‘when it happens’ problem and not an ‘if it happens’ problem.”
Similarly, Little urges schools to start the education process now. “No one expects the process to be perfect,” she says. “Just take the first step. Students are watching and acknowledge when you step out and stand on principle in alignment with the values of your school.”