Ask Korii Notice for Parents

Version: v1.3
Last updated: 2026-11-06 · Effective date: 2026-11-07

This English text is the governing version. Consent records for Ask Korii refer to this notice as korii-notice-v1.2.

This notice is for parents and guardians. Please read it before you switch on “Ask Korii” for a child. It supplements our Privacy Policy and Children's Privacy Notice.

1. What Ask Korii is

Korii is our mascot. “Ask Korii” lets a child type questions about learning Chinese characters (for example “How do you write 山?”), which are answered by AI. Your child is talking to an AI system, not a person. Answers are generated automatically by AI, may be wrong, and do not replace a teacher. Questions unrelated to learning characters receive a fixed reply.

2. Your separate consent

Ask Korii is off by default. After reading this notice, you switch it on for each child separately in the parent area. We record that you consented, the version of this notice (korii-notice-v1.2) and the time. You can switch it off at any time, with immediate effect.

3. How your child's questions are handled

• Each question is first checked on our own servers. Inappropriate words, questions unrelated to learning characters, and anything about being hurt or wanting to hurt oneself receive a fixed reply from us and are never sent to the AI.
• If a question contains names, schools, school years, addresses, phone numbers or similar, we remove them before anything is sent to the AI, and remind the child that Korii does not need to know these things.
• Other questions, together with information about the character from our own data, are sent to Anthropic's Claude API (in the United States) to generate an answer. We do not send your child's nickname, your email address, account or any identifier.
• Each answer is checked again on our servers (for example, it may not invent characters or pronunciations) before your child sees it.
• We do not keep the wording of your child's questions or of the answers.

4. What we record

• How many questions each child has asked that day, to apply the limit of 20 a day: only the number, never the content; deleted after 2 days.
• Daily totals (how many questions were answered, how many received a fixed reply, how often each kind of safety reply was shown), not linked to any child or account, to check quality and safety; kept for 12 months.

5. Fixed safety replies

• If a child mentions feeling sad, being hurt or wanting to hurt themselves, Korii does not let the AI answer. It shows a pre-written message encouraging the child to tell a parent, teacher or trusted adult straight away, with a helpline for their region where we have checked one against an official source (currently Malaysia and New Zealand).
• If a child mentions hurting someone else, Korii likewise does not let the AI answer; it shows a pre-written message encouraging the child to talk to a trusted adult right away.
• If a child mentions someone asking for contact details, wanting to meet, or asking them to keep a secret, Korii reminds the child to tell a parent or teacher.
• We do not notify you or keep your child's words because of these replies.

6. How Anthropic treats the questions

Anthropic generates answers only on our instructions and does not use this content to train its models. It deletes it within 30 days by default, or may keep it for up to 2 years if its safety systems flag a Usage Policy violation. No one other than Anthropic receives what your child types.

7. Languages and limits

Ask Korii explains in Chinese, English, Malay and Arabic. Malay and Arabic answers are first written in English and then translated by the same AI service. Each child can ask up to 20 questions a day. It is currently a free trial and may sometimes be unavailable.

8. Your rights and contact

You can switch Ask Korii off at any time, and you can view, export or delete your child's information as set out in the Privacy Policy. Questions: privacy@chuschool.com.