Is AI safe for kids to use?

Short answer: mostly yes, with supervision — and the risks aren't the ones most parents are worried about.

We teach kids 8–14 to build games and apps using AI. That means we've watched a lot of children use these tools, and we've watched a lot of parents worry about it. The worry is reasonable. The specific things people worry about are usually not the things that actually matter.

What parents usually ask about

The most common question is whether the AI will show a child something inappropriate. The major tools — Claude, ChatGPT, Gemini — all have content filtering, and in the context your child will be using them for, this almost never comes up. A kid asking "write me a script that makes a coin disappear when a player touches it" is not wandering into dangerous territory.

The second question is about data. What is the AI learning about my child? This one is worth taking seriously, and it has a practical answer: don't let your child put personal information into an AI chat. Not their full name, not their school, not their address. This is the same rule you'd apply to any website, and it's worth saying out loud rather than assuming they know.

Both of the major tools also have age requirements in their terms — typically 13, sometimes 18 with parental permission. That's why every account setup in our books is marked as a step to do with a grown-up. The account should be yours.

The risk nobody mentions

The real risk isn't that AI harms your child. It's that your child learns to accept whatever it says.

AI is confidently wrong on a regular basis. It will write code that doesn't run. It will explain something incorrectly with total conviction. A kid who learns to paste whatever the AI produces and hope for the best has learned nothing useful — and has picked up a habit that will hurt them for years.

This is why we build our whole approach around judgement rather than output. The skill isn't getting the AI to make something. The skill is looking at what it made, noticing it's wrong, and knowing what to ask next. Every project in our books ends with the child testing what they built and fixing what broke. That loop is the actual lesson.

What good supervision looks like

•     Set the accounts up yourself, using your email.

•     Keep it in a shared space — not behind a bedroom door.

•     Ask what they're building. You don't need to understand the code to follow the project.

•     Make "is that actually right?" a normal question, not an accusation.

•     Handle anything involving uploads, purchases, or publishing yourself.

The comparison that matters

The honest way to think about this isn't "is AI safe" in the abstract. It's: compared to what?

Your child is likely already spending time on a device. The question is whether that time is spent consuming something or building something. A kid using AI to make a game is doing something structurally different from a kid watching videos of other people making games — even though from across the room, both look like a child staring at a screen.

We're not going to tell you screen time is fine because it's educational. We'll say this instead: the kid who finishes a project has something to show for the hours. That's a real difference, and it's the one we build for.