AI safety rules for kids: a family agreement

Rules a child helped write are followed. Rules delivered as a list are negotiated around. Treat this as a conversation you have once and revisit occasionally, not a policy you post on the fridge and enforce.

The six rules

• We give the AI ideas, not facts about us. No names, addresses, schools, or photos.
• A grown-up knows when we're using it. Not watching every second: just knowing.
• If something feels weird, we say so straight away and nobody gets in trouble.
• The AI can be confidently wrong. We check anything that sounds like a fact.
• We don't put other people into it: not friends, not siblings, not teachers.
• We make things with it. We don't ask it to be our friend.

How to introduce them

Ask your child what they think the rules should be first. Children are frequently stricter than their parents, and a rule they proposed is one they will remind you about later. Add whatever they missed rather than starting from your list.

Rule three is the important one

"Nobody gets in trouble" is the rule that determines whether the other five ever get used. A child who fears consequences hides the moment something went wrong, which is precisely the moment you need to know about. Say it out loud, and then honour it the first time it is tested.

Revisit at each new tool

The rules are the same; the risks are not. A drawing tool and a chatbot warrant different levels of attention, and a child moving from seven to ten can carry more of the reasoning themselves.