AI and child safety

What the AI does

With your child's work, two things only, and only if a parent gives permission. It reads handwriting: a photo of an answer, or a worksheet page a parent sends, is turned into text. And it marks answers the answer key cannot mark: it compares a child's answer or working with the question's marking scheme, or uses its own judgement where there is no answer key. Its marks are final and labelled “Marked by AI”. That is the whole feature for your child's work.

Separately, the Tuition Guru content team also uses AI to help draft study guides from our own material. No child's work, name or details are ever sent for this. A person on our team checks every guide and must approve it before any family sees it, and it is labelled when AI helped draft it.

What it will never do

• Your child never chats with the AI. There is no chatbot, and children cannot send anything to it themselves.
• It never gives your child a mark unless you have given the AI permission. With it, AI marks are final when given and say “Marked by AI”; you can question any mark in one line, it is re-checked automatically, and you can then ask a person to look, once. Without it, no AI mark is given and the answer waits.
• It does not diagnose, assess ability, or comment on a child's behaviour, emotions or learning needs, and it builds no profile of them.
• It does not address or reply to your child, and it never claims to be a person or a teacher.
• It does not ask for personal information. Where a name or similar detail is visible, it is instructed to leave it out.
• It is never used to make a decision about your child.

An adult is always in control

• Only a parent or guardian can switch on AI reading and marking.
• Permission is separate, per child, and can be withdrawn at any time without losing anything that does not use AI.
• You black out anything you do not want shared and approve the redacted copy before it is sent.
• Every AI mark says “Marked by AI”. What the AI reads from a worksheet page you send stays a draft until you have checked it.
• The AI can be wrong. Where it is unsure, a second, independent AI checks the mark before it is given, and your child gets the higher mark. You can question any mark, and ask a person to look.

How we hold the AI to this

Every request carries a written, versioned child-safety instruction setting out the rules above, and the version is recorded with the result. If the reply does not match the strict format we expect, it is discarded rather than shown. Anthropic is our external AI processor; we do not use its chat, file-storage or feedback features, and we never submit your child's work to it in ratings or bug reports.

If age-inappropriate content appears in an image, the AI is instructed to refuse it and not describe it.

Raising a concern

If something the AI produced was unsafe or upsetting, or you are worried about a child's wellbeing, tell us through Report a safety concern when signed in. A person reads every report, and your child's work is never sent to the AI as part of handling one.

If a child is in immediate danger, call the Police on 999. To report suspected child abuse in Singapore, call the National Anti-Violence and Sexual Harassment Helpline on 1800-777-0000, open 24 hours. This service is not an emergency service.

What we do not claim

Tuition Guru holds no safety certification, accreditation or regulatory approval, and we do not claim any endorsement by Anthropic. We do not publish an accuracy figure: our testing so far uses made-up practice pages, which are easier to read than a real child's handwriting, so any number from it would mislead you. This is why an adult reviews everything.