An AI Framework for Parents and Kids: Raising Capable, Careful Children in the Age of AI

1. What AI is good for, and the four risks that matter

A child can use AI to practise a language, hear a hard idea explained three different ways, get unstuck on homework or make something they are proud of. The risks are ordinary parenting problems in a new form:

• Accuracy. A chatbot states wrong answers with full confidence.
• Exposure. It can surface content a child is not ready for.
• Privacy. It keeps whatever personal information a child types in.
• Relationships. For some children it starts to take the place of a friend.

You have handled versions of all four with television, phones and the front door. The same calm footing works here.

The guidance from UNESCO, Common Sense Media and school frameworks points the same way. Children do better with AI when a parent is nearby, when the device is shared and when the family has talked about it. A parent who has tried the tool and keeps asking about it gives more protection than any filter.

2. The framework: three questions and a ladder

Ask three questions about any moment your child wants to use AI. Is the thinking theirs? Is their private data staying private? Is it taking the place of a real person?

The second half is a ladder. A five-year-old and a fifteen-year-old need different rules, so think in age bands, each adding a rung of independence the child has earned.

3. What to do at each age

Under 7: parent-led play, no account. Use AI together, for fun. Ask an image tool to draw the silliest animal your child can invent, or let them dictate a bedtime story and laugh at what comes back. The lesson is that grown-ups control the computer and stay in the room.

Ages 7 to 10: side by side on a shared device. Your child can follow their own questions with you next to them. Start the habit of doubt: when the tool answers, ask out loud whether it is true and how you could check. A child who learns early that AI is confident and sometimes wrong carries that into every later use.

Ages 11 to 13: supervised practice with written rules. Homework is where habits set. AI may check work, explain what the child did not understand and quiz them before a test. It may not produce work they hand in as their own. Keep the device in a common area and, where the service allows, keep the account under a parent's email.

Ages 14 and up: guided independence. Your role moves from gatekeeper to coach. Talk about the harder points: a chatbot can sound like a friend without being one, anything typed in is not private, and skipping the thinking now shows up at exam time. Teenagers hide what they expect to be punished for, so honest conversation protects them better than tight control.

The tool below shows the recommended posture for each age band. Treat it as a prompt for a conversation. You know your child better than any framework.

4. Household rules you can agree tonight

Children keep rules they helped write. Sit down together and agree a short list; here is a starter set to edit.

• AI lives in the open. We use it on shared devices in shared spaces, not alone behind a closed door.
• We never tell it who we are. No full name, school, address, photo or phone number goes into a chatbot.
• We check before we trust. If it matters, we confirm what AI says with a real source or a real adult.
• It helps us think; the work stays ours. AI can check our work and explain things. It does not do the work we are meant to learn from.
• People come first. When something is bothering us, we talk to a person we trust.
• We tell a grown-up if it gets weird. If AI says something scary, mean or strange, we come and tell you, and no one is in trouble for asking.

5. Four guardrails and how to set each one

Accuracy. Ask the tool something you already know the answer to, with your child beside you, and let them catch it being wrong. One real mistake they spot themselves teaches more than any lecture.

Exposure. Switch on the parental controls and teen modes the major services now offer, and keep use in shared spaces. No filter catches everything, so the backstop is a child who will tell you what they saw.

Privacy. Teach one line until it is automatic: the chatbot is a stranger, and we do not tell strangers private things. Turn it into a game of spotting what should never be typed in.

Independence. Watch what AI is replacing rather than counting minutes. If it is replacing a friend, a hard conversation or the struggle of working a problem out, steer back. Children learn most in the stretch where a task is hard but possible.

The second tool sorts common uses into green, amber and red. Tap a use to see why it lands where it does, and use it to build your family agreement.

6. Signs the framework is working

Measure it by how your child uses AI when you are not watching:

• They catch it in a mistake and tell you.
• They stop before typing a real name or school.
• They bring you something odd a chatbot said, expecting help.
• They go to a friend, not a screen, when they are upset.

The part of this framework I am least sure of is the teenage band. Fourteen-year-olds vary enormously, and the research on older teens and AI companions is still thin. Adjust that rung faster than the others if your child shows you they can handle it, or cannot.

"We are not raising children to survive AI. We are raising them to use it well, and to stay more human than the machine while they do." - Adrian Dunkley, AI Boss

What to do this week

1. You, tonight: open the AI tool your child uses most and spend 15 minutes asking it questions you know the answers to. You cannot guide what you have not tried.
2. You and your child, this weekend: agree the six household rules above, edit them together and stick the list near the shared device.
3. You, in account settings: check each AI service's minimum age and turn on any parental or teen controls. Where the service allows it, put accounts for under-14s on your email.
4. You, with your child's teacher: ask what the school's AI policy says about homework, so home and school draw the same line.
5. Once a month: ask your child to show you one thing AI got wrong and one thing it helped with. Five minutes keeps the conversation open.

Frequently Asked Questions

At what age can my child legally have their own AI account?

Most major AI services set 13 as the minimum and several require parental consent up to 18. Those limits follow data laws: in the United States, COPPA restricts collecting data from children under 13, and under the EU's GDPR the default age of digital consent is 16, which member states can lower to 13. Check each service's terms, because the minimum can differ by country.

How do I set up parental controls on an AI chatbot?

Look in the account settings for "parental controls", "family" or "teen" options. Several major services, including ChatGPT since 2025, let a parent link to a teen's account and adjust features such as memory, image generation or quiet hours. Where no controls exist, use the device's own controls to limit which apps a child can install.

What should I do if my child has already used AI to write homework?

Ask them to explain the work to you out loud before deciding anything. If they cannot, the issue is learning, not just rules, and the fix is to redo the task with AI only allowed for explanations. Talk to the teacher about the school's policy, and agree the homework line together so it does not feel like a trap.

Are AI detectors a reliable way to catch AI-written homework?

No. Detection tools produce false positives, and a 2023 Stanford study found they flagged essays by non-native English writers as AI-written far more often than essays by native speakers. If a school accuses your child on detector evidence alone, ask to see drafts, notes or version history instead, and keep those records as a habit.

How can I tell if my child is leaning on a chatbot for emotional support?

Watch for secrecy about chats, talking about the bot as if it were a person, and choosing it over friends or family when upset. Companion apps designed to act as friends or partners are a different category from homework tools, so check app store age ratings and what is installed. If you see these signs, talk calmly and, if your child seems low, involve their doctor or school counsellor.

Further reading, and the full course

This framework follows the guidance schools and researchers use, translated for the dinner table. For the primary sources, start here.

• UNESCO guidance on generative AI in education, which recommends a minimum age of 13 for independent use and a human-centred approach. unesco.org
• Common Sense Media research and family guidance on how children use AI, and how to keep it safe at home. commonsensemedia.org
• The school frameworks your child's teachers may be using, including TeachAI and the Australian Framework for Generative AI in Schools, so home and school can say the same thing.