Little learners.
Real people. Clear boundaries.

Learning has a purpose.

We use AI for learning explanations, voice interaction and some story creation. Activities also include authored questions, answers and hints. A finished activity records practice; it is not, by itself, proof that a child has mastered an idea.

We encourage children to explain an idea to a grown-up and try it away from the screen. A short, useful learning moment is a good place to stop.

Warmth, with an honest introduction.

Our conversation instructions tell the AI to be clear that it is a computer learning tool, not a person. It should not claim human feelings, ask for a secret relationship, say it needs a child, or pressure them to stay. If a child asks whether it is real or wants a close personal relationship, the guide explains what it is and points back to people and learning.

You can explain it like this: “Shiku is a pretend character powered by a computer. We can use it to practise. It can get things wrong, and we can ask a real person for help.”

Sensitive topics need real support.

Kiddora checks conversation text for certain unsafe requests and personal details before asking its AI provider for an answer. It also checks returned answers. These checks and the AI’s instructions can miss things or misunderstand what a child meant.

For recognised danger or distress, the response encourages help from a trusted adult. A child can choose a safe adult, such as a parent, teacher or another caregiver. Kiddora cannot assess emergencies or replace support from a person.

Voice is a family choice.

Voice-enabled child conversations use the parent account’s voice setting. The device also asks for microphone permission. Tapping and reading remain available without a microphone.

The green listening wave appears after recording starts, reacts to microphone volume, and disappears when recording stops. Each recording is limited to 20 seconds and can stop sooner after silence. The Stop microphone control ends the turn; leaving the page or hiding it also stops recording.

Short recordings are sent to the speech service to recognise words. Kiddora’s speech endpoint does not save those recordings or transcripts; service-provider handling is described in our privacy information. Do not share addresses, phone numbers, passwords or other private details.

Different voices. More than one way to try.

Speech recognition can mishear accents, quiet voices and speech differences. A missed word is not a judgement of a child’s ability. Use a tap, read with a grown-up, or try again when it feels comfortable. We do not claim equal recognition accuracy for every child.

We need to be clear about the limits.

AI can give an incorrect or inappropriate answer despite safeguards. Generated stories need a grown-up’s review before sharing. General explanations do not replace a qualified teacher’s judgement or medical, legal or emergency help.

This page describes the current product, not a guarantee that every response is safe or correct. Last updated 11 October 2026.