Family AI Agreement Template

Companion to write a family AI agreement your kids actually helped write. Draft this together in one or more age-appropriate conversations, with the child present — not announced after the fact. Participation does not transfer safeguarding responsibility to the child.

Household: ____________________________ Date drafted: ____________________

1. Accounts

Whose account, for what, at what supervision level:

Current product minimum ages checked (date-stamp this):

• OpenAI ChatGPT: ____________________
• Anthropic Claude: ____________________
• Google Gemini (Family Link): ____________________
• Other: ____________________

2. Private data and approved exceptions

Information never entered into a consumer AI tool:

• Other children’s identities
• Home or school location
• Credentials or account-recovery details
• Intimate images
• Health or disability details
• Another person’s information without authority to share it

Any school-approved exception must record its purpose, exact tool/account, minimum necessary data, consent or policy basis, retention rule and responsible adult. “It is for school” is not itself approval:

3. Schoolwork

Our household’s method: the homework AI-use decision tree.

The specific teacher’s or school’s rule always overrides this agreement’s default. Agreed: Yes

4. Companions

After checking current age rules and qualified child-safety guidance, our conservative household rule for persona-based companion apps is:

• Not permitted for [child’s name]

Reasoning discussed together: ____________________________________________

5. Creation

Permitted: ____________________________________________

Never permitted:

• Images of real people without their consent
• Content depicting classmates or family members in situations that didn’t happen
• Content intended to deceive someone (fake screenshots, manipulated quotes, synthetic voice)

6. Purchases

No purchase, subscription, or in-app payment an AI tool suggests happens without specific adult approval — every time, no exceptions.

Agreed: Yes

7. Help-seeking

If something goes wrong — a frightening conversation, an off-feeling request from a stranger, an image appearing somewhere it shouldn’t, or a mistake the child is afraid to admit — coming to an adult leads first to support and safety. Adults should not promise outcomes they cannot control; danger, exploitation, sexual content, threats or extortion come before any later rule discussion.

Adult(s) to go to: ____________________________________________

Signatures

Parent/guardian: ____________________________ Date: ____________________

Child: ____________________________ Date: ____________________

Review date

Next scheduled review, chosen for this child and product context: ____________________

Revise earlier if: a new AI product enters the household, a violation reveals ambiguous wording, or a real circumstance changes (new device, sibling reaching a new age band, school policy change).