Child safety online

Every AI character on Questipedia runs behind a curriculum-scoped system prompt and a content filter. They are instructed to refuse off-topic personal chat, romance, self-harm, or anything unsafe for children — and to redirect the pupil to a trusted adult when they detect distress.

AI answers are grounded in curriculum content and do not attempt to profile the child.

Pupils cannot be direct-messaged privately by adults on the platform. Adult-to-pupil contact only occurs inside supervised, recordable lesson contexts with a verified educator. Peer chat is scoped to classrooms, houses, and squads that a school or parent has authorised.

• Automated filters for profanity, grooming keywords, self-harm signals, and PII sharing.
• Rate limits on messaging to prevent spam and coordinated abuse.
• Escalation to human review for any flagged incident.
• Suspension of accounts pending review where risk is high.

Any pupil, parent, or educator can raise a report from any page using the flag icon. Reports create a ticket in the admin queue with the original message, timestamp, and context preserved.

Serious safeguarding concerns are triaged within 24 hours (7 days a week). Referrals to the police, LADO, or the NSPCC helpline (0808 800 5000) are made where required by law or professional duty.

Where our moderation stack detects self-harm or safeguarding language, the pupil is shown Childline (0800 1111) and Samaritans (116 123) signposting immediately, regardless of whether a human has yet reviewed the incident.

• Talk to the pupil about what to do if something online makes them uncomfortable.
• Use the parent account (or the school's designated adult) to review activity summaries.
• Report anything unsafe promptly using the flag icon or by emailing our safeguarding inbox.