Responsible AI

KAERON uses AI in exactly one place a student can see: when your child writes an answer to a Try it or Challenge task, an AI model reads that answer and replies with a score and a few sentences of encouragement about what to try next.

That is the whole of it. In particular:

• There is no AI chatbot or AI tutor a student can talk to freely. We have deliberately not built one.
• AI does not decide whether your child passes, progresses or earns a certificate.
• AI does not write the curriculum. Every lesson, activity and project was written by a person.
• AI is never used to profile your child, to target advertising, or to make any decision about them outside the lesson they are working on.

The lessons teach AI as a subject — what it is, how it works, where it is useful and where it fails. That is reading and building, not an AI system acting on your child.

When an answer is scored, the request to the AI provider contains four things and nothing else:

• the title of the lesson,
• the task your child was asked to do,
• the note the lesson author wrote about what a good answer shows, and
• the text your child wrote.

Your child's name, email address, school, class, age and account identifier are not sent. The model is given a piece of schoolwork to mark and is told the writer is a child aged roughly 10 to 14 so that its tone is right. It is not told who they are, and it has no way to find out.

Because of this, a written answer is the one piece of your child's work that leaves our systems. We tell students plainly, in the lessons themselves, not to type personal details — their full name, address, school, phone number or anyone else's — into an answer box. If you spot something in your child's answer that should not be there, tell us and we will remove it.

We use OpenRouter, a service that routes a request to one of several underlying model providers. Our request is made from our server, never from the page your child is looking at — the key that authorises it exists only on the server and is not present in the website's code in any form.

This matters for more than secrecy. Because every AI request passes through our own backend, we control what is in it, we can change provider without touching your child's experience, and we can switch the feature off entirely if a provider stops meeting the standard on this page.

Being finalised. OpenRouter forwards a request to whichever underlying model answers, so the country a given answer is processed in can vary. We are pinning down the specific list of providers we permit and where each one processes text, and will publish both here alongside links to their terms.

An AI model does not know things. It predicts likely text. It will say something false in a confident voice, and it will do it often enough that treating its output as authority is the single worst habit a student can pick up from it.

So the programme teaches the opposite habit:

• Check it. If AI states a fact, find it somewhere that can be checked.
• Run it. If AI writes code, run the code. Code that looks right and does not work is the most common result.
• Understand it. Work your child cannot explain is not their work yet, whoever typed it.
• Say when you used it. Using AI is allowed. Passing its work off as your own is not.

The score a student receives is a piece of feedback from a machine that can be wrong too. A student who thinks their answer was misjudged can say so, and a mentor will look at it.

The grader is deliberately built so that it can fail without costing your child anything. Their answer is saved before any AI is contacted. If the provider is slow, broken, or switched off, the answer sits waiting to be marked by a mentor instead — the work is never lost and never silently discarded.

A mentor can review and override any AI score. If you disagree with feedback your child received, write to us through the contact page and a person will look at the actual answer.

Consent for your child to use the AI-assisted parts of the programme is asked for separately at checkout, as its own tick — not bundled into accepting the terms. You can withdraw it by writing to us, and we will confirm in writing what changed.

Withdrawing it means written answers are marked by a mentor rather than scored automatically. Feedback may take longer to arrive. Nothing else about the course changes, and no part of the curriculum becomes unavailable.

Being finalised. How long a graded answer and its feedback are kept is part of the retention schedule being finalised for the privacy policy. Until it is published, ask us and we will tell you what we hold about your child and remove anything you ask us to.