In the first 48 hours after a deepfake of a student surfaces, a principal should do four things in this order: stop anyone looking at, copying or forwarding the image; put the child's welfare in the hands of the safeguarding lead; name who holds the record, the removal requests and the message; and decide nothing about the alleged maker until the child is safe.
Take a hypothetical case. At 7:40 on a Monday morning a teacher shows you a phone. A fabricated sexual image of a fifteen-year-old pupil, built from her photo on the school website, is moving through a group chat. She made nothing and sent nothing. The next hour will shape the next year.
How many schools are ready for this?
Few are, and many do not know whether they are. An Education Week survey of 852 educators, reported on November 4, 2026, found that one in three did not know whether their school or district has a policy on AI-generated fake images and videos, and 22% said theirs has none. Nine in 10 told the EdWeek Research Center they had not been trained on detecting or responding to deepfakes in the last two years. The survey is American; the page does not give field dates or a margin of error.
The English picture is not calmer. A survey of 500 UK headteachers by the safeguarding company Aidos, reported by Schools Week in September 2026, found that 49% had experienced an incident of pupil images being misused, manipulated or used in an extortion attempt, and that only 43% said their image consent process addressed the risk of deepfake generation. It is a vendor survey, so read it as a signal rather than a census.
Children are also reporting it themselves. The Internet Watch Foundation said on August 10, 2026 that its Report Remove service received 397 reports in 2025, and 420 in the first half of 2026, in which the child believed at least some of the imagery was faked. "Faked" there is the child's belief, not an IWF finding. The direction of travel is clear enough.
What is the Four Holders rule?
The Four Holders rule is how I suggest a school gives a deepfake incident an owner on day one: four jobs, four named people, none of them the same person. One holds the child, one holds the record, one holds the removal, and one holds the message. An incident where one person holds all four will be handled slowly. An incident where nobody holds the record is one the school cannot later explain.
This is my suggested way of thinking, not a validated method. It exists because of a pattern in how these incidents go wrong, which I come to below.
| Holder | What they hold | What they never do |
|---|---|---|
| The child | Welfare, safety, one trusted adult, contact with the family. Usually the designated safeguarding lead (or in a US school, the counselor or administrator who leads student safety) | Investigate who made the image |
| The record | Who reported what, when, where it was seen, and who was told. Words and times only | Store, copy or screenshot the image itself |
| The removal | Requests to platforms and services, and the line to police or children's services where the law requires it | Contact the alleged maker or demand deletion from their phone |
| The message | What staff, families and the board or governors are told, and when. Usually the principal | Say more than is known |
The principal should hold the least. That surprises people. Holding the message and the decisions only a principal can take is already a full job on the first day.
What must nobody do in the first hour?
Nobody should look at the image again, copy it, forward it or keep it. Four mistakes are common, and each is made by someone trying to help.
The first is the screenshot "for evidence." In England, the government's guidance on sharing nudes and semi-nudes (last updated March 11, 2024) says that computer-generated "pseudo-images" which look like photographs are treated as indecent images, and its summary of the viewing rule is that staff should avoid viewing unless there is good and clear reason, and should never copy, print or store the images. In the US, the Education Week report quotes Erin Mote of InnovateEDU warning that a teacher who asks a student to send a deepfake intimate image to a personal phone then holds child sexual abuse material on it. The rule underneath is the same in both countries: the adult's job is to know the image exists, not to hold it.
The second mistake is the confrontation. A principal who summons the suspected maker that morning has acted before the child has a single protective adult around her, and has possibly tipped off the person with the files. The third is the group warning, an email to the year group that spreads the very thing it condemns. The fourth is deciding the response is a discipline matter. Center for Democracy and Technology policy counsel Kristin Woelfel makes the point in the same Education Week report: detection matters less than response, because the harm to the child is the same whether the image is real or fake.
What does the 48-hour card look like?
It looks like four windows, each with a named holder. Copy it, put names against it this week, and keep it where a deputy can find it at 7:40 on a Monday.
| Window | What happens | Holder |
|---|---|---|
| First hour | Stop viewing and sharing. Secure the device without copying anything. Tell the safeguarding lead. Name the four holders. Where police or children's services must be told, make the call | Child, record, removal |
| By hour four | Contact the pupil's family unless doing so would put her at risk. Offer a named adult and a place to be. Start removal requests. Write down every action with its time | Child, removal, record |
| By end of day | A short, factual note to staff: what has happened, what they must not do, whom to tell. Wording for families agreed with police or children's services if they are involved | Message |
| By 48 hours | Removal requests chased. The chair of governors or board told. Next steps for the alleged maker decided under your behavior policy, after safeguarding, with restorative practice considered alongside sanctions | Message, removal |
In England, the same guidance says incidents should always go to the police through the Multi-Agency Safeguarding Hub where there are aggravating factors such as an adult involved, coercion or blackmail, or a child at immediate risk. Its summary says families should be informed and involved early unless that would put the child at risk. In the US, mandated reporting duties and school-police arrangements vary by state and district, so the removal holder needs your district counsel's number before the day it matters. I am not a lawyer, and nothing here is legal advice.
Can the image actually be taken down?
Sometimes, and quickly if the right route is used. In the US, the federal Take It Down Act was signed on May 19, 2025. The Federal Trade Commission began enforcing its platform provisions on May 19, 2026: a covered platform must remove a reported intimate image, including a digital forgery, and known identical copies, within 48 hours of a valid request. The duty falls on platforms, not schools, so the school's role is to make sure the family knows the route exists and to help them use it. The law firm Fisher Phillips, in a June 2025 briefing, also advises schools to check their conduct policies and to keep careful documentation of deepfake incidents.
In the UK, an under-18 can self-report nude or sexual imagery of themselves, including imagery they believe is faked, through the IWF and Childline's Report Remove service, and the IWF can work to remove and block illegal imagery. The removal holder should tell the child about it, but the child has to choose to use it.
Removal is slower where the image sits in a private messaging group, which is why the first hour's work on stopping the spread matters more than any later request.
What I See in Practice
The mistake I see most often is that the school has a plan for bullying, a plan for online safety and a plan for data breaches, and the deepfake falls through all three. Across the leadership teams I work with, the telling sign is who answers when I ask who would hold the record. The usual answer is the safeguarding lead, who is also the person holding the child, and who cannot do both well in the same hour. What helps, I would argue, is not a longer policy. It is a decision made in advance about who does what, and a conversation with each of those people.
As a school trustee, I know how boards tend to work: they ask whether a policy exists. The better question is whether the four people can be named without looking anything up. The questions I would put to a head are in What Should a Chair of Governors Ask the Head About AI?, and the board-reporting side is in What Should You Report to Governors About AI?.
What if the target is a teacher?
The same four holders apply, with one change: the person who holds the child now holds a colleague, and the record should hold more, because a staff member may have legal options a pupil does not. Do not let a teacher's embarrassment push the incident down the list. The Education Week report names teachers and students as targets, and the response a school gives a teacher tells every other teacher what to expect.
What should you settle before it happens?
Five things, and none needs a new policy.
- Name the four holders on one page, with a deputy for each, and put the page where a deputy can find it.
- Add AI-generated and AI-altered images to your nudes and semi-nudes or intimate-image procedure by name, so that no one has to decide in the moment whether a fake counts. Keeping Children Safe in Education 2026, in force in England from September 1, 2026, describes such images as ones that "may also be digitally altered or wholly generated using artificial intelligence."
- Remove identifiable pupil photos from public pages, or review who can save them, and make the image consent form say what the school will and will not publish. The Aidos survey suggests many forms are silent on this.
- Rehearse the first hour once a year, with the phone in your hand and a stopwatch running.
- Agree, in advance, what a pupil who made an image will face, and how that is separated in time from the protection of the child. The wider question of what a school does when a student is accused is covered in What Should a School Do When a Student Is Accused of Using AI?.
For the safeguarding changes that sit around this, What Should a Designated Safeguarding Lead Change Because of AI? and What Does AI Mean for Safeguarding in Schools? take the same ground from the lead's and the institution's side, and Who Is Accountable When AI Gets It Wrong? covers the question of who answers for the outcome.
The Leadership Question
The question for a principal is whether, on the worst morning of the year, the four people already know they are the four people. If your leadership team is working through that, this is the kind of work I support through AI strategy sessions and advisory work for schools.
Sources and further reading
- Banerji, O., Fake AI Images Are Targeting Students and Teachers. Schools Aren't Prepared, Education Week, November 4, 2026 (US; EdWeek Research Center survey of 852 educators).
- Schools Week, Heads ask for more help on how to conquer misuse of pupil images, September 2026 (England; Aidos survey of 500 headteachers).
- Internet Watch Foundation, Hundreds of British under-18s reporting nude or sexual imagery of themselves say it's faked or manipulated, August 10, 2026.
- Federal Trade Commission, Take It Down Act enforcement starts now, May 19, 2026.
- Fisher Phillips, New Federal AI Deepfake Law Takes Effect: 4 Steps Schools Must Take, June 3, 2025.
- UK Department for Education and UKCIS, Sharing nudes and semi-nudes: advice for education settings, GOV.UK, last updated March 11, 2024.
- Department for Education, Keeping children safe in education 2026, GOV.UK, in force September 1, 2026.
Dan Fitzpatrick is the founder of The AI Educator, a Forbes contributor and an international keynote speaker who helps school and system leaders lead through AI.


