Your family’s one-page AI agreement

Most families do not need another lecture about AI. They need one calm page everyone helped write before the next homework question, health worry, polished message or too-convincing phone call appears.

That page can hang on the fridge. It is not a contract in the legal sense, and it will not stop every bad decision. Its power is simpler: it turns “Be careful” into a shared pause. What may we paste? When do we check? Who do we ask? What do we never hand over, even when we are tired, rushed or flattered by a beautifully written answer?

A recent survey of about a thousand Australian parents who already use AI regularly found the same tension in many homes: 58% put the privacy and security of family data among their top concerns, and 47% worried about accuracy, yet only 42% said they review or discuss how their children use AI. Because these parents were already AI users, the figures do not describe every family. Still, they capture a familiar gap. Concern is high. Conversation is occasional. Children—and adults—are left to improvise in private.

An agreement narrows that gap. It gives the family a common language before a specific situation makes everyone defensive.

Why writing it together works

Rules handed down from adults often become a test of enforcement: Who saw? Who can prove it? What counts as cheating, oversharing or disobedience? Rules made together work differently.

When children help choose the boundaries, they are more likely to remember them and to use them with friends, siblings and even adults. A child can say, “That’s on our never-paste list,” without making the moment a personal accusation. Adults also become easier to trust when the page applies to them. The fastest way to make an AI agreement feel like surveillance is to exempt the people who wrote it.

Signing is partly theatre, and that is not a weakness. A signature says: This is ours. We chose it. We can revisit it. It also makes the page a household reference, not a speech someone must deliver again under pressure.

Start without a family cold war

Do not begin after an incident, late at night, or with a printed list of prohibitions. Pick twenty unhurried minutes and bring a blank page.

Start with a generous question: “What has AI actually helped us with this month?” Then ask each person to offer one green use, one worry and one rule they would like applied to somebody else. The last part matters. Children should propose at least one rule for the grown-ups—for example, no asking AI to write an apology without reading it aloud first, or no uploading family photos for a funny edit without everyone’s okay.

Expect disagreement. Write the lowest-risk version you can all accept, and let the review date settle the rest. You are not deciding forever; you are running a household experiment.

For younger children, keep the spoken version to three ideas: ask first, private stays private, and check important answers. Older children can help word the fuller page and decide which examples belong beside each rule. With teenagers, be explicit that this is not permission to inspect every chat. The agreement governs risky moments and honesty when they arise—not constant monitoring.

Copy, adapt and sign this version

OUR FAMILY AI AGREEMENT

Family members: [names] Start date: [date] Review date: [date]

1. GREEN LIST — AI IS WELCOME TO HELP WITH Brainstorming ideas; explaining something in a simpler way; practising questions; organising notes we wrote; suggesting recipes or activities; and helping with a first draft after we have made our own attempt.

For schoolwork, AI can coach, quiz and explain. It does not replace [name]'s first try, reading, calculations or final judgement.

2. NEVER-PASTE LIST — WE DO NOT PUT THIS INTO AI Full names combined with identifying details; home or school addresses; school name with a class timetable; passwords or codes; identity, medical, financial or legal documents; photos of documents; private messages; or personal information, images and stories about other people without their permission.

If a document matters, we describe the problem, remove names and numbers, or ask a trusted adult/professional to look at the original.

3. ASK A GROWN-UP FIRST IF [child name], age [age], will pause and ask [adult names] when: - a person, app or AI asks for personal information, money, a photo, a password or secrecy; - an answer concerns health, money, safety, law, discipline or a big family decision and it feels wrong, surprising or too certain; - an image, voice or video appears to show a real person; - AI suggests contacting, accusing, paying, meeting or sending something to someone; or - the result will be handed in, published, posted or shared.

Asking is not getting in trouble. It is the agreement working.

4. OUR TWO-STEP ACCURACY HABIT For anything important, we take two steps: Step 1: Ask, “What is this based on, and what might be missing?” Step 2: Check one key claim against a second trusted source—such as a teacher, book, official organisation, professional or known website.

AI can help us find what to check. It is not the final check.

5. IF A CALL OR MESSAGE FEELS STRANGE Our family safeword is: [safeword] If a voice or video request is urgent, involves money or secrecy, or refuses the safeword, we hang up and contact the person another way.

6. THESE RULES APPLY TO EVERYONE Adults follow the never-paste list too. We say when AI helped with an important message. We do not secretly check private chats unless there is a serious safety concern. If someone reports a mistake early, we fix the problem before we argue about blame.

Signed: [signature / name] [signature / name] [signature / name] [signature / name] ```

Our separate guide to AI voice scams explains the wider playbook; the safeword above is the one line your family needs on the fridge.

After it goes on the fridge

During the first few weeks, refer to sections rather than motives. “Green list or ask-first?” is easier to hear than “You knew better.” If a rule is broken, notice what made it hard to follow. Was the wording vague? Was a child embarrassed to ask? Did an adult paste a document because the safer route took too long? Rewrite the friction, not just the child.

Protect the norm that early honesty gets help. A child who says, “I sent this and now I’m worried,” has done exactly what the page was designed to encourage. So has an adult who admits, “I believed that answer too quickly.” Those moments teach more than perfect compliance.

Then keep the expiry date. AI tools, school expectations and your children’s independence will change faster than a permanent rulebook can. Every school term or long holiday, read the page together and ask three questions: What helped? What felt unnecessary? What new use needs a decision? Cross things out in pen. Initial the changes. A visibly revised agreement is not a failed agreement—it is a family learning in public.

The goal is not to make AI frightening, or to pretend one page can make it safe. It is to remove the silent guesswork. When everyone knows what is welcome, what stays private, when to pause and how to check, AI becomes easier to use well—and much easier to talk about before something goes wrong.