Protecting Your Child's Data When Using AI Tools

Quick answer

What actually happens to your child's information when they use an AI tool for homework? In most cases: an account email is stored, usage patterns are logged, and everything typed into a conversation is kept by the company behind the tool, and may be used to improve future versions unless a setting is switched off. None of this needs to alarm you, but it does need sensible handling. The two habits that matter most are simple: keep identifying details - full name, school, health information - out of prompts altogether, and set up accounts with the same care you would apply to any other service holding your child's data. Get those two things right and the rest is manageable detail.

What AI tools actually collect

Every mainstream AI tool collects roughly the same categories of information: the email or phone number used to sign up, basic device and usage data, and - the part most people forget - the full content of every conversation. That last category matters most for a teenager, because a chat about a maths problem can drift into personal territory without anyone intending it to: a mention of a family situation, a health worry, a friend's private drama.

Most major providers state clearly, in their privacy policies, that conversations may be used to train and improve future versions of the tool unless a user opts out. This is not unusual or sinister by the standards of modern software, but it is worth knowing plainly rather than assuming a chat is somehow private the way a diary is.

Why this deserves specific attention for a teenager

A teenager typing quickly, thinking out loud, is not weighing every sentence the way an adult drafting a formal document would. That is exactly the moment identifying or sensitive details slip in - a full name next to a school and suburb, a passing mention of a medical appointment, a family financial detail dropped in while asking AI to help plan a budget.

None of this is really about a single bad prompt causing harm. It is about the cumulative picture that builds up in an account over months of ordinary use, and how much detail is really necessary for that account to still be useful. The habits below solve for exactly this, and they sit alongside the wider set of habits covered in AI education for teenagers in Australia. For the broader safety picture beyond privacy specifically, see Is AI safe for teenagers?

Account hygiene that actually helps

What should never go into a prompt

Some details simply do not belong in an AI conversation, regardless of how safe the tool otherwise is. Full name alongside school and suburb. Anything medical or health-related, whether about your teenager or a family member. Passwords or account numbers, even hypothetically, "for an example". Anything a friend told them in confidence. A useful rule of thumb for your teenager: if they would not want a stranger reading it back to them, it should not go into the prompt.

This matters more, not less, as AI tools get better at holding long-running memory of a user across conversations. A detail typed once can resurface later in ways that feel unexpectedly personal.

Training-data opt-outs are real, and worth using

It surprises many parents that a training-data opt-out exists at all, but on the major tools - ChatGPT, Claude, Gemini and similar services - it does, usually as a setting inside account or privacy preferences. Switching it on does not stop the tool working; it simply stops that account's conversations being used to improve future versions of the model. It is one of the few genuinely useful privacy controls available, and it is worth five minutes to find and switch on for any account your teenager uses regularly.

The Australian Framework for Generative AI in Schools treats this territory seriously too: Privacy, Security & Safety is one of its six guiding principles, which is a reasonable signal that the habits above are not overcaution, they are simply what good practice already looks like.

Common mistakes parents make

• Assuming AI chats are private, when in most cases they are stored and sometimes used for training by default.
• Letting an account carry a full name, school and suburb, when a first name is usually enough.
• Reusing an existing personal password, which turns one compromised account into several.
• Never checking the privacy or training settings, when the useful controls are often a single toggle away.
• Treating this as a one-off setup task, when a short review each school term keeps the habit current as tools change.

The recommendation: set the account up properly once, with a first name only, a dedicated email, and the training opt-out switched on, then reinforce one simple rule for every prompt afterward: nothing identifying, nothing medical, nothing that would embarrass your teenager if a stranger read it. That combination of a sensible account and a sensible habit covers the overwhelming majority of the real risk, without turning every AI conversation into an anxious exercise.