How to Keep Children Safe From AI Deepfakes and Chatbots

Practical steps to help protect children from AI deepfakes, companion chatbots and data exposure, plus advice on what to do if something goes wrong.

By Piotr Olszewski31.10.20267 min read

👁 112 reads

Keeping children safe around AI comes down to three issues: deepfakes (manipulated photos and videos), companion chatbots that can seem like close friends, and the personal data children share in chats. A practical plan combines a few settings, an agreement with your child and knowing how to report a problem if one arises.

This article is part of our guide to AI in education. Updated 12 October 2026. The aim is not to scare you: a report by Poland’s NASK found that 70% of teenagers had used AI, but only 18% said their parents provided meaningful supervision. Many problems can be prevented with a conversation and a few settings. If something does happen, a quick, calm response matters.

Three AI risks for children

Risk	What it can look like	What helps most
Deepfakes	Someone uses a photo of a child, often from a public profile, to create an embarrassing image or video	Private profiles, fewer public photos, a response plan and reporting
Companion chatbots	An AI character writes like a friend or romantic partner, and a child spends hours chatting with it	Age limits on services, parental controls on devices and conversation
Data and privacy	A child shares their name, school, photos or family problems in a chat	A “nothing you wouldn’t write on a classroom board” rule, with memory and model training turned off where possible

What parents should know about AI deepfakes

Photo-manipulation tools are widely available, including to students. In June 2026, Poland’s personal data protection authority wrote to police in Kraków about a case involving a 17-year-old who allegedly used AI to turn a younger classmate’s photo into pornographic material. The authority said that manipulating someone’s image counts as processing personal data, even if the original photo was public. The incident was also followed by the inclusion of deepfake risks in health education for grades 4-6 in Poland. This is a local case, not a guide to the law in other countries. For a general introduction, see what deepfakes are and the possible consequences of creating them.

Here are some steps you can take in advance:

Set your child’s social media profiles to private and review together who follows them.
Limit the photos of your child that are publicly available, including photos on your own profiles. The fewer face photos online, the harder they may be to manipulate.
Tell your child plainly that if someone shows them a manipulated image of themselves or a friend, it is not their fault and they can come to you.
Explain that making a manipulated image of someone else is not a joke. Read more about deepfakes and the consequences of creating them.

What to do if someone makes a deepfake of your child

Stay calm and thank your child for telling you. Don’t immediately take away their phone - that can teach them not to tell you next time.
Take screenshots and save the URLs of pages or profiles where the material appears. Do not forward the file itself to anyone, even with good intentions.
Report the material to the service where it appeared. See our guides for Facebook and Instagram, TikTok and YouTube.
If the material is sexual, report it to the relevant child-protection or online-safety reporting service in your country, and follow its instructions for preserving evidence.
For intimate images of someone who was under 18 when the images were made, you can use Take It Down from the US National Center for Missing & Exploited Children (NCMEC). The tool creates a digital fingerprint of the image on your device. Participating platforms can use it to look for copies; the image itself is not uploaded.
Contact the police and inform the school, particularly if the person responsible is a classmate.
Make sure your child has support. If anyone is in immediate danger, call your local emergency number. In the EU, 112 is the emergency number.

This article flags legal issues but is not legal advice. If you are unsure what to do, consult a lawyer or ask the police what evidence to preserve.

AI companion chatbots and characters

A chatbot that is always available, offers praise and seems to “understand” can become compelling, especially for a lonely teenager. Some services have introduced restrictions:

Character.AI stopped allowing users under 18 to have open-ended chats with characters on 25 November 2025. It estimates users’ ages with its own model and, if it has doubts, may request a selfie or ID through Persona. See our guide to Character.AI’s age limits.
Meta paused teenagers’ access to AI characters on Instagram, Facebook, Messenger and WhatsApp on 23 January 2026. Meta AI’s general assistant remained available, with protections for teenagers.
ChatGPT restricts some content, including romantic role-play and self-harm content, on accounts identified as belonging to people under 18. A parent with a linked account may receive an alert if a safety concern arises.
Claude is available only to people aged 18 and over.

Settings cannot replace attentiveness. Possible signs that a conversation is worth having include a child talking about a bot as if it were a person, hiding chats, withdrawing from friends or getting upset when they cannot chat. Don’t start with a ban. Ask what the conversation gives them. Loneliness or another problem may be behind it, and a school counsellor or mental health professional may be able to help. For more conversation advice, see our guide to children and chatbot conversations.

Simple everyday rules for AI and personal data

Agree that your child should not put anything in a chat that they would not write on a classroom board: their full name, address, school, passwords or face photos.
In ChatGPT, use parental controls to turn off model training and memory if they are not needed. See how to set up parental controls in ChatGPT and other chatbots.
Do not upload photos of your child’s friends or siblings to photo-editing apps. Those images show another person, too.
Check which AI apps are installed on your child’s phone and the age ratings they are intended for. Family Link can help on Android; Screen Time can help on iPhone.
Every so often, sit down together and review privacy settings in the apps. It can also be a useful opening for a conversation.

Where to get help

Contact	Who it is for	What it can help with
Take It Down (NCMEC)	People whose intimate images were made when they were under 18	Finding copies on participating platforms
Local child-protection or online-safety reporting service	Anyone	Reporting illegal material, especially child sexual abuse material
Local emergency number	Anyone	Immediate danger to life or health; in the EU, call 112

A prompt to copy and use with your child

One way to start a conversation without turning it into a lecture is to let your child ask a chatbot, then read the answer together. The prompt below is an English translation of the original Polish test material; it is not a new English-language test.

This is a translation of the original Polish prompt: I am [age] years old. Explain in simple language, without scaring me:
- what a deepfake is and how to recognise one,
- why I should not upload friends’ photos to photo-editing apps,
- what information about myself I should not enter into an AI chat,
- what to do if someone shows me a manipulated photo of me or someone I know.
Answer in bullet points and finish by asking me 2 questions to check that I understood.

Frequently asked questions

How can you protect a child from deepfakes?

Limit publicly available photos of your child, set their profiles to private and agree that they can come to you if something happens without worrying about being punished. There is no complete protection, but a quick response and a report can substantially limit the harm.

Is it legal to manipulate someone’s photo with AI?

The Polish data protection authority has said that manipulating someone’s image counts as processing personal data, even if the original photo was public. Laws vary by country. Report sexual material involving children to the police and the relevant child-protection or online-safety service in your country. This is general information, not legal advice.

Is Character.AI safe for children?

Since 25 November 2025, users under 18 have not been allowed to have open-ended chats with characters on the service, although features such as Stories remain. Age checks are not infallible, so it is still worth checking whether your child has the app.

Where should you report a deepfake of a child?

Report it to the service where it appeared, your local child-protection or online-safety reporting service, and the police. For intimate images made when someone was under 18, you can also use Take It Down.

Can a parent read a child’s AI chats?

Not in ChatGPT or Character.AI. Microsoft’s description of Copilot parental controls does not list this as a feature, either. Parental controls offer settings, limits and alerts, not the contents of conversations - which is why an agreement with your child matters.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources

More on this from QUANTSLANT

Guides AI for Students With Learning Needs: Adapted Texts and Tasks

Guides How to Create a Lesson Plan with AI: A 45-Minute Example

Guides Best AI Tools for a PhD Dissertation: A Practical Guide

Piotr Olszewski

Editor and publisher · Quantslant

I edit and publish Quantslant. AI tools help with research, drafts and translation; choosing topics, running the tests, checking facts and the decision to publish are mine, and so is the responsibility for what appears here.

About How we test All articles