Skip to content
Curiotide

What Kids Should Never Share With AI: A Parent's Checklist

  • All ages
Illustration of a golden padlock floating above calm teal ocean waves

Kids tell chatbots things. That’s not a character flaw; a friendly voice that asks follow-up questions is practically an invitation. The problem is on the other end: what goes into an AI tool may be stored, reviewed, or used to train future models, and once it’s gone in, you can’t pull it back out. So this is the practical version of “be careful”: a concrete list, a game that teaches it, and the settings that back you up.

Why this matters (the 60-second version)

Three things can happen to what your child types or says into an AI app:

  1. It can be stored. Many tools keep conversation history on company servers by default.
  2. It can be used for training. Some services use conversations to improve their models unless you opt out.
  3. It can leak. Companies have breaches. Data that doesn’t exist can’t be breached.

None of this requires a villain. It’s just how the plumbing works, which is why the rule is about what goes in, not about trusting or distrusting any particular company. The FTC’s children’s privacy rules (COPPA) limit what services may collect from under-13s, but rules are a floor, not a guarantee.

The never-share list

Put this next to the ground rules on the fridge. Kids should never give an AI tool:

  • Full name (first name alone is usually fine)
  • Home address, or enough detail to find it (“the blue house next to the gas station on…”)
  • School name or teacher names
  • Photos or videos of themselves, siblings, or friends
  • Phone numbers or emails, theirs or anyone’s
  • Passwords, ever, for anything
  • Family information like where parents work, daily schedules, or when the house is empty
  • Money details they’ve glimpsed: card numbers, account anything
  • Big secrets and heavy feelings. Not because they’re data, but because those belong with people who love them. (More on that in AI is a tool, not a friend.)

The park-stranger rule

A list is easy to forget. This rule of thumb isn’t: if a friendly stranger at the park could know it, it’s fine to share. If not, it stays private.

A stranger at the park can know your first name, that you love dinosaurs, and that you’re seven. They cannot know your last name, your school, or your address. Kids can apply this test on their own, which is the whole point: privacy rules only work when they function without you standing there.

The sorting game (five minutes, weirdly fun)

Say a fact, and your child shouts “SHARE” or “PRIVATE”:

  • “I love pizza” (share)
  • “I go to Lincoln Elementary” (private)
  • “My dog is named Biscuit” (share)
  • “My mom’s phone number is…” (private)
  • “I’m scared of thunderstorms” (trick one: fine for the park, but big feelings still go to people)

Play it in the car. Let your child make up facts to stump you. Two rounds of this teaches more than any lecture, and you’ll find kids proudly correcting each other within the week.

Grown-ups leak too

A gentle mirror check: the checklist applies to us. Pasting your child’s full name and diagnosis into a chatbot to draft a school email, or uploading family photos to an image tool without checking how they’re used, shares their information just as surely. Model the rule you’re teaching, and narrate it: “I’m changing your name to ‘my son’ before I ask it this.”

Settings that back the rule up

Habits first, settings second. But do take ten minutes to:

  • Turn off chat history or model training where the tool allows it (many chatbots have a setting like “don’t use my conversations for training”).
  • Use kid or education modes where they exist; they typically collect less. Our tools roundup favors tools built this way.
  • Check what’s already stored in any tool your family uses, and delete old conversations while you’re there.
  • Review app permissions: an AI app rarely needs your child’s location, contacts, or photo library.

Common Sense Media’s privacy program maintains privacy evaluations of popular kids’ apps if you want a second opinion on a specific tool.

What to do if something already went in

Don’t spiral, and don’t scold; your child telling you is the win (that’s ground rule 7). Delete the conversation if the tool allows it, use the account’s data-deletion option if one exists, and treat it as the world’s most effective teachable moment. Then play the sorting game again at dinner.

The goal is a kid whose privacy filter runs on its own, without you in the room. A stranger at the park does that job.

Keep reading