Nudify apps and deepfakes: the conversation to have before it happens

Short answer: have the emergency conversation tonight, before anything happens, and teach one rule that covers almost everything: nothing goes online, into an app or to a friend that you would not hand to a stranger in the street. The tools that turn an ordinary photo into a sexual fake are free, fast and being used in schools now. Your child cannot stop that from existing. What you can do is make sure that if it touches them, they come to you within the hour, not within the year.

Here is what the tools do, what the numbers say, the words I use, and the steps to take if it happens.

What the tools do

I am not going to name apps or explain how they work, because the how is not the point and the apps change monthly. The shape of it is enough.

A "nudify" tool takes a clothed photo of a real person, a school photo, a holiday snap, a profile picture, and generates a realistic fake in which that person is undressed. A face-swap or deepfake tool puts a real face onto someone else's body in a photo or video, or makes a person appear to say things they never said, in their own cloned voice. Two years ago this needed skill and hours. Now it needs a screenshot and a minute, and the tools have been advertised in the same feeds your child scrolls.

In the school cases I hear about, the person in the image is usually a girl and the person who made it is usually a boy she knows. And the image, once made, travels through the class WhatsApp group faster than any adult can follow it.

The numbers

Chapter 18 of the book gives the figures I trust, so I will use them exactly.

In a 2024 RAND survey, 13% of school principals reported incidents of students using AI deepfakes to bully or harass peers within the previous two school years. That is one principal in eight, for a technology that barely existed in schools two years earlier.

The Internet Watch Foundation's data for 2024 documented 245 actionable reports containing 7,644 confirmed AI-generated child sexual abuse images, an increase of around 380% on the year before, and in an earlier one-month snapshot found more than 20,000 AI-generated images posted to a single dark-web forum. Some of those images start as ordinary photos of ordinary children.

The conversation to have before it happens

Every family has the "what to do in a fire" talk long before there is smoke. This is the same talk. I have it in a calm moment, in the car or over the washing-up, never straight after a news story, and I say three things.

One: the stranger rule. "Anything you put online, into an app or send to a friend, assume a stranger could end up holding it. So only send what you'd be fine showing a stranger in the street." This is the rule Chapter 18 gives for AI tools, and it covers photos sent to a boyfriend, selfies in a group chat and pictures uploaded to a filter app, because any of those can be the source photo for a fake. I say clearly that a fake does not need a nude to start from. A school photo is enough. That is why the rule is about everything, not about "sexting".

Two: it is a crime, even for a teenager, even as a joke. This is the sentence most children have not heard. Here in South Africa the Cybercrimes Act of 2020 makes it a criminal offence to disclose an intimate image of someone without their consent, and the law's definition includes simulated images, so a sexual deepfake counts. The Films and Publications Act separately prohibits distributing private sexual images to cause harm. And any sexual image of a person under 18, real or generated, is child sexual abuse material, a serious crime for whoever makes it or shares it, including a 15-year-old who forwards it "so you can see how bad it is". In May 2025 a Pietermaritzburg court sentenced a man to five years for cyber forgery and disclosing intimate images. I am a teacher, not a lawyer, and the details differ by country, but the direction is the same almost everywhere: making it, sending it and forwarding it are all the offence. Chapter 11's Offline-Online Bridge makes this land: "How would you feel if someone did that to you? Even if you know it's a joke, could it be used to humiliate you?"

Three: you will never be in trouble for telling me. This is the emergency conversation from Chapter 18, and I use the book's words almost exactly: "If you ever see something online that upsets you, a picture, a video, a message, or if someone makes you uncomfortable, I'm always the right person to come to. I won't punish you. I won't take your phone away. I'll just help you deal with it." The reason this matters more than any filter is that early disclosure is the variable that most decides how these situations end. A fake that is reported within an hour is in three phones. A fake that is reported after a month is in three hundred.

How to spot a deepfake, for teenagers

Give your teenager the tells from Chapters 11 and 18, and give them honestly.

Look at the edges. Unnatural blinking, or none. Strange shadows that fall the wrong way. Lighting on the face that does not match the room. Backgrounds that are impossible, or blurred where they should be sharp. Mismatched details: an earring on one side only, a hand with the wrong number of fingers, hair that merges into a collar. In video, a mouth that does not quite match the audio.

Then look away from the image. Chapter 11's lateral reading is the stronger skill: do not stare at the picture, open new tabs and investigate where it came from. Who posted it first? Does a reverse image search find the original photo? Google's "About this image" feature can show when an image first appeared online and where. If a shocking image of a classmate has no source except a screenshot in a group chat, that absence is the tell.

Then check your own pulse. Emotional scepticism, also from Chapter 11: content built to make you furious, disgusted or frightened is exactly the content to slow down on. Chapter 18's Pause and Check habit puts a number on it: thirty seconds before sharing anything surprising or emotionally charged.

And the honest note: the visual tells are fading. The best fakes now have the right number of fingers. Within your child's teenage years, looking harder will stop working, and asking "where did this come from?" will be the only reliable question. Teach the tells, but teach source-checking as the skill that lasts.

What to do if it happens

If your child comes to you, or you find out another way, this is the order I would follow.

Stay calm and do not punish. Your child, whether target or bystander, needs to see that telling you was the right decision. Even if they sent the original photo. Especially then. The lecture can wait a month; the next steps cannot wait an hour.

Preserve the evidence without spreading it. Screenshot the image, the messages around it, the usernames, the dates and the links, and keep them on one device. Do not forward the image to anyone, including the school, the other parents or the police, because forwarding it is itself distribution and, if the child in it is under 18, a crime. Describe it; hand over the screenshots in person if asked.

Report to the platform. Every major platform has a route for reporting intimate images shared without consent, and increasingly they treat AI fakes under the same rule. Report every copy you can find.

Tell the school the same day. Ask for the principal or the designated safeguarding person, in writing, and ask what their procedure is. Your child's school is where the image is moving and where the other children are.

Report to the police. In South Africa, any SAPS station can take the report, and the Family Violence, Child Protection and Sexual Offences unit deals with these cases. Take your screenshots and your notes. Elsewhere, the equivalent is your local police and, in many countries, a dedicated online-harm hotline.

Use the removal services. The Internet Watch Foundation takes reports of sexual images of under-18s and works to get them taken down. Take It Down, run by the US National Center for Missing and Exploited Children, lets a young person, or a parent on their behalf, create a digital fingerprint of an image without uploading it, so participating platforms can block it. Both are free.

Look after the child. The fake is not real; the humiliation is. Say that sentence out loud. Get the school counsellor or your GP involved early, and keep the child out of the group chats, not out of school.

By age

Ages 5 to 8. No talk about nudify apps; the concepts are wrong for the age. Two things instead. First, body privacy in the ordinary way: which parts are private, and that photos of private parts are never okay for anyone to take. Second, the seed of media scepticism: "Some pictures on a screen are made up by a computer. Can you tell which?" Play it with obviously silly AI images, and start the emergency conversation in its youngest form: "If anything on a screen makes you feel funny, you tell me, and you'll never be in trouble."

Ages 9 to 12. This is when phones and group chats arrive, so this is when the stranger rule goes in, along with Chapter 18's privacy rule that no photo showing your school, your street or your face goes into a general AI tool. Add the first version of sentence two: "Sending on a private picture of someone is against the law, even if you're a kid, even if someone sent it to you first." Use Chapter 8's "The Screenshot" dinner scenario, where everyone in the group chat is laughing at a mean screenshot of a classmate, because the skill of not forwarding is learned on small things. Have the emergency conversation in full.

Ages 13 to 18. Say all three sentences, plainly, including the law and the word "crime". Teach the tells and the lateral reading, and share the RAND figure with them as data, not a scare. Talk about the boyfriend or girlfriend case directly: a photo sent in trust is one breakup away from a fake, and asking for one is not love. Write it into the Family AI Agreement from Chapter 10: what never gets uploaded, what never gets forwarded, and the Uh-Oh plan for who you tell first. The free builder has a place for it.

Questions parents ask

What is a nudify app?

A tool that takes an ordinary clothed photo of a real person and generates a realistic fake in which they appear undressed. They are cheap or free, fast, advertised on social media, and are being used by school-age children against classmates. Making or sharing such an image of anyone under 18 is child sexual abuse material and a serious crime, including for a teenager.

Is making a deepfake of a classmate illegal in South Africa?

If it is sexual, yes. The Cybercrimes Act of 2020 criminalises disclosing an intimate image without consent and its definition covers simulated images, so a sexual deepfake counts; the Films and Publications Act also prohibits distributing private sexual images to cause harm. Any sexual image of a person under 18, real or generated, is a serious crime for whoever makes or shares it, whatever their age.

How can I tell if a photo or video is a deepfake?

Look for unnatural blinking, shadows and lighting that do not match, impossible backgrounds, mismatched small details and lips out of sync with the audio. Then do the more reliable check: search sideways for where the image first appeared, using reverse image search or Google's "About this image". If the only source is a screenshot in a group chat, treat it as fake until proven otherwise.

What should I do if someone made a deepfake of my child?

Stay calm and do not punish. Screenshot the image, messages, usernames and dates, and forward the image to nobody. Report it to the platform, tell the school in writing the same day, report to the police (in South Africa, any SAPS station or the FCS unit), and use the Internet Watch Foundation and Take It Down to get it removed. Get your child support early.

Sources

• Chapter 18 of Raising AI-Ready Children on safety and epistemic risks, the "only upload what you'd show a stranger" rule, the emergency conversation and the Pause and Check habit.
• Chapter 11 on deepfakes and digital manipulation, lateral reading, emotional scepticism and the "It's Just a Joke" red flag.
• Chapter 8 on the ethical reasoning scenarios by age; Chapter 10 on the Family AI Agreement.
• RAND Corporation, 2024, "Artificially Intelligent Bullies: Dealing with Deepfakes in K-12 Schools".
• Internet Watch Foundation, Annual Data and Insights Report 2024, on AI-generated child sexual abuse imagery.
• Northwestern University, Kellogg School of Management, "Detect Fakes" research on human detection of manipulated media.
• Cybercrimes Act 19 of 2020 (section 16) and the Films and Publications Act, South Africa; the Pietermaritzburg sentencing reported in May 2025.
• Internet Watch Foundation reporting portal and Take It Down (National Center for Missing and Exploited Children), checked on the date at the top of this post.