Your little one asks for one more story, and the tablet is ready. Before you say yes, a quiet question comes up: is AI safe for children?
This guide shows you what to watch for, and how to keep things calm at home.
Is AI safe for children? Start with the kind of tool you’re choosing
You may be wondering whether AI is safe for your child. That worry is completely understandable. The honest answer depends on three things: what the tool does, how old your child is, and how close an adult stays. A bedtime story generator is very different from an open chatbot or an AI companion.
Not every AI tool carries the same risk
Open chatbots can answer almost anything, which is why they need watching. They may invent facts, repeat stereotypes, or produce material that is frightening or unsuitable for a small child. Voice assistants bring their own concerns: recordings made by accident, unintended purchases, and conversations that sound a little too human.
Educational AI tends to stay narrow. It is built around one subject or one skill. That focus can limit what your child sees, especially if a teacher has checked the tool. Even so, an adult should stay nearby, because a confident answer can still be wrong.
A young child may be ready for short activities shared with an adult nearby. They should be able to tell you what the tool is doing, stop when you ask, and bring anything confusing to you. Independent conversations need much stronger judgment, and that usually comes later.
Before your child tries a tool
- Can your child tell you what the tool is doing?
- Will your child stop when you ask?
- Will your child bring anything confusing straight to you?
Children can meet false answers and emotional traps
AI can make mistakes that look polished and certain. Your child might receive a wrong homework explanation, an invented fact, unsafe health advice, or a story that slips in violence or adult themes. Encourage your child to pause and check anything important with you, a teacher, a book, or a trusted official website.
Children are already turning to AI for personal worries. In research across ten countries, about one child in ten who used AI said they had asked it for advice about something that troubled them.
Some children come to see AI as a companion or a comfort. That can feel soothing in a lonely moment. But a chatbot cannot truly understand feelings or reliably notice danger.
AI can also be used to create scams, fake news, and realistic sexual deepfakes built from a child’s photo or video.
Teach your child not to type names, addresses, school details, passwords, medical history, or private information about a friend. Photos, videos, and voice recordings need extra care, because they can reveal faces, homes, uniforms, locations, or other people. If anyone threatens or pressures your child, tell them to come to you straight away.
AI can support learning, but it should not do the thinking
AI tutoring can help with a specific skill when it offers hints, practice, and feedback instead of simply handing over the answer. Learning gains are generally positive, but smaller than those from ordinary tutoring, and less clear over longer periods.
Personalised practice can meet a child where they are
A focused learning tool can adjust the difficulty, repeat a tricky idea, change the wording, or offer another example. That can help a child practise reading, spelling, vocabulary, maths, or a new language at a pace that feels comfortable.
Instant help has a soft edge, though. When you are busy, a child may accept the first answer just because it arrived so quickly. Encourage them to ask, “How do you know?” and “Can you show me another way?” before they trust the result.
Keep room for effort, mistakes, and understanding
Some struggle is genuinely useful. Before opening an AI tool, ask your child to try one step, sketch what they already know, or explain the problem out loud. The tool can then offer a hint instead of taking over the work.
In France, supervised and explained use of generative AI can support classroom learning from quatrième, roughly age 13, when it follows the curriculum and the teacher’s rules.
Using AI to complete all or part of homework without the teacher’s permission, and without your child’s own work, can count as fraud.
With a preschool child, keep the explanation simple: “This is a computer tool that makes words from patterns. It doesn’t know things the way people do, and it can make mistakes.” Then ask your child to explain the answer back to you in their own words.
Age limits vary by service and do not prove suitability
An age limit tells you who can use a service. It does not tell you if the chats, images, memories, ads, or design suit your child. So read the feature rules, not just the account rule.
Popular services set different age rules
Minimum ages and conditions differ from one service to another:
- ChatGPT: not intended for children under 13. Users aged 13 to 18 need parental permission, and OpenAI advises that an adult run the interaction when ChatGPT is used educationally with a child under 13.
- Gemini Apps: the rules depend on the account and the country. In France, which is part of the European Economic Area, Gemini Apps with a Google AI plan require users to be at least 18. Supervised Family Link accounts in the European Economic Area cannot currently access Gemini Apps.
- Microsoft Copilot: generally allows access from age 13, although regional rules and individual features can change that threshold.
- Character.AI: lists a minimum age of 16 in Europe and has introduced separate age checks and teen protections.
French consent rules do not set a service’s minimum age
In France, children under 15 usually need consent from a parent, as well as their own, when an online service relies on consent to process personal data. This rule covers data processing. It is not a general permission to use every AI service.
If an account is truly eligible, enter the correct age. Choose the child or family account option. Review the privacy settings together. Keep recovery details under adult control.
Never get around an age rule by entering a false birthday. A false date can leave your child without the protections built for younger users.
Protect privacy by sharing less personal information
AI services may collect your account details, your prompts, uploaded files, images, voice recordings, device information, and usage data. What they collect depends on the service and on the settings you choose.
Keep identifying details out of prompts
Your child should not type:
- a full name
- an address or phone number
- the name of their school
- their daily route
- passwords, medical history, or family finances
- private information about a friend
A harmless-looking story can reveal a lot once these details are put together.
Photos, videos, and voice recordings need extra care. They can show a child’s face, home, uniform, location, accent, or habits. They can also show other people who never agreed to be uploaded. When you can, use an invented character or a cropped image that identifies no one.
Explain privacy in calm, simple words. You could say: “What you type may be stored, checked, or used to improve the service. So we only share what we’d be comfortable putting on a postcard.” Your child needs an explanation they can understand, not a long legal document.
Before your child types a prompt
- Replace any real name with an invented character
- Leave out the school, the address and the daily route
- Crop photos so no face, home or uniform shows
- Keep passwords and family money matters out of the chat
Check storage, training, and deletion controls
ChatGPT lets account holders manage several privacy controls:
- Model improvement, which can be turned off
- Temporary Chat
- Deleting conversations
- Exporting data
- Deleting the account
Temporary chats don’t appear in history or train models while they stay temporary. They may still be kept for up to 30 days for safety.
Gemini offers activity controls, deletion options, auto-delete periods, and temporary chats.
Some conversations reviewed by human reviewers can be retained for up to three years, so avoid entering a child’s sensitive information even when a privacy switch is available.
Adult supervision works best with clear, regularly revisited boundaries
Start with a first session together. Pick one simple task, watch how the tool responds, and ask your child what felt helpful or strange. Show them how to stop, report a problem, leave, or ask an adult for help.
Make family rules easy to remember
Keep the rules short and visible:
- Purpose: AI for a named activity, such as practising a word or making a story, rather than endless browsing.
- Place: AI stays in a shared room or beside an adult, not behind a closed door.
- People: serious feelings, frightening messages, bullying, or threats go straight to a trusted adult.
Use device-level screen limits and age filters where they are available. Service controls can also reduce sensitive content or set quiet hours. Filters are imperfect, and secret monitoring can weaken trust, so explain what you check and why. These tools cannot replace your presence.
Controls that help versus controls that hurt
Helpful
Clear family agreements, proportionate checks, and honest explanations of what you can see and why.
Harmful
Secret monitoring, which may weaken trust and push children to hide what they do.
Watch the child, not only the screen
Revisit the arrangement every few weeks. Ask what your child uses AI for. Ask whether an answer upset them, whether they feel pressured to keep going, and whether they would rather talk to the tool than to a person.
Take these signs seriously:
- Secrecy
- Sudden distress
- Anxiety when access stops
- Late-night use
- Withdrawal from play
- Repeated requests for personal information
These signs call for a calm conversation, not blame. If AI has been used to dodge homework, look for the difficulty underneath. Your child may feel lost, rushed, ashamed, or afraid of getting things wrong.
Talk with the teacher. Then break the task into smaller steps. The chatbot is rarely the only problem.
If the tool gives an unsafe answer, save the relevant details, close the conversation, and tell your child they did the right thing by showing you.
For self-harm, abuse, sexual exploitation, or immediate danger, contact emergency services or an appropriate child protection professional.
AI should not replace real relationships or screen-free routines
AI companions can make friendship seem easy. They are always there, always agreeable, and never disagree. Real relationships are harder. Children learn to listen, to repair hurt, to wait, to compromise, and to handle feelings that don’t match their own. A screen cannot teach them that.
Watch for warning signs. If AI starts to crowd out friends, imaginative play, sport, hobbies, family conversation, or sleep, it is time to step in.
Protect the day before screens start
- Set a stopping point before use begins.
- Keep devices out of the bedroom.
- Protect parts of the day that belong only to people.
For children under five, the World Health Organization recommends no sedentary screen time at age one. At age two, it recommends no more than one hour a day, and less is better. Active play, sleep, reading, and storytelling with a caregiver matter far more than another screen activity.
If you use a service such as TellMeStories, share only the details the story needs. Check the text before you read it. Never upload a photo or voice recording that could identify your child. Read the story together, and stay the warm centre of bedtime.
Keep the first question in mind, because it will change as your child grows. A tool that is fine for a five-year-old practising letters may need a different answer at thirteen, and again at sixteen. Check the rules each time a new app appears on the screen, and make a habit of asking your child what they used it for today.
Frequently Asked Questions
Can AI tell me whether my child has a mental health condition?
No, a chatbot cannot diagnose a child or reliably judge how serious a symptom is.
If an answer worries you, talk with your child, your doctor, or a school professional. In immediate danger, contact emergency services.
Should my child use an AI tool alone while travelling?
For a young child, avoid unsupervised use during travel. A tool can suddenly produce something confusing, frightening, or unsuitable, so prepare a shared device, offline activities, or content you have checked beforehand.
Prepare a shared device, offline activities, or content you have checked beforehand, so you stay in control.
What should I do if my child shares a private image with AI?
Stay calm, stop any further sharing, and keep a record of the relevant account details.
Report the content through the service. If the image is sexual, threatening, or being used to pressure your child, contact emergency services or an appropriate child protection professional for urgent help.
How often should I review an AI service?
Review it whenever the service adds a feature, changes its age rules, or your child starts using it for something new.
Add a short check-in every month to catch changes in privacy, content, habits, or emotional reliance early.
