Ask most schools what worries them about AI and it isn't the technology. It's that pupils are already using it, fluently and daily, without anyone having taught them how to do it well. Teaching children to use AI safely is not about restricting access they already have. It is about building the judgement to use it responsibly. This post sets out what that looks like in practice, at both KS2 and KS3.
Start with what pupils already do
There is little point teaching AI safety as though pupils have never met it. Most have. They use it to draft, to summarise, to answer questions and to make images. The honest starting point is to acknowledge that and build on it: not "don't use AI", which pupils will ignore, but "you're using it, so let's make sure you're using it well".
This framing matters because it keeps the conversation open. A pupil who has been told AI is banned will use it in secret and learn nothing about using it safely. A pupil taught to use it thoughtfully will tell you how they used it, which is exactly the behaviour you want.
The habits worth teaching
Safe AI use comes down to a small set of habits that transfer across any tool.
The first is to question the output, because AI can be confidently and fluently wrong, and pupils need to learn that a well-written answer is not the same as a correct one. The second is to check before trusting, treating AI as a starting point to verify rather than a final authority. The third is to protect personal information, understanding what should never be typed into a tool. The fourth, and most important, is to keep doing their own thinking, using AI to support learning rather than to skip it.
Taught well, these are not rules to memorise but instincts pupils carry into any tool they meet later.
What this looks like at KS2 and KS3
At KS2, keep it concrete. Younger pupils can grasp that a computer can produce something that looks real but is not, that it can make mistakes, and that some information is private. The emphasis is curiosity paired with caution.
At KS3, the same ideas gain depth. Pupils can explore why AI gets things wrong, how bias enters a model, and when using AI on a piece of work is appropriate and when it defeats the purpose. This is the age to move from "AI can be wrong" to "here is why, and here is how I would check".
Teaching it as a structured progression, rather than repeating the basics each year, means each stage builds on the last.
Where safeguarding fits
Teaching safe AI use overlaps with online safety, and the two are strongest taught together. Pupils need to know not just how to use AI, but what to be wary of: what to share, what not to, and when something they have seen or been asked feels wrong. Framing AI within the wider picture of being safe online helps pupils see it as part of how they behave, not a separate technical topic.
Making it stick
The difference between a school that has "done AI safety" and one where it has actually landed is evidence. A one-off assembly is easy to deliver and easy to forget. A structured approach, revisited across the year with a way to see what pupils have understood, is what turns a message into a habit. It also lets a school show, rather than assert, what its pupils can do when a parent or inspector asks.
That evidence layer is what we built Curio to provide: a structured AI literacy programme across KS2 and KS3, with a record of what each pupil has learnt. If it would help to see how it works, we'd be glad to talk.