NorAI
For curious kids, ages 6–14
An AI friend
you actually
control.
Your child talks. You decide how it thinks, how it talks back, when it's available — and you can read every word.
I'm an AI buddy, not a real person
Hi Maya!What shall we wonder about?
Hold to talk
What's in it
Everything is built around one idea: the fun belongs to your child, the control belongs to you.
Voice first
Kids hold one big button and just talk. No typing, no spelling, no keyboard — which matters most for the youngest ones who have the best questions and the worst spelling.
Guardrails on both sides
Every message is checked before it reaches the AI, and every reply is checked before your child sees or hears it. Self-harm language is answered instantly and calmly, without waiting on the model.
You see everything
A dashboard with the full conversation log, a flagged-only filter, daily usage, and an audit trail of every safety decision the system made and why.
You set the hours
Available hours per day of the week and a daily minute budget. Outside them your child gets a gentle 'come back later' screen, not an error.
One profile per child
Each child gets their own name, age, interests, reading level, language, time limits and conversation history — independent from their siblings.
Their own buddy
They name it and pick its face. It knows their interests and talks at their reading level, so a six-year-old and a twelve-year-old get genuinely different companions.
How the safety actually works
Most kid-safe AI is one instruction to the model and a hope. A prompt is a request, not a guarantee — so we built the part that doesn't depend on the model behaving.
- 1
Before the AI hears it
Your child's words are screened for self-harm, profanity, grooming patterns, and personal information — in English and French — including disguised spellings kids actually try.
- 2
Crisis, answered instantly
Self-harm language never reaches the model. It gets a calm, immediate reply pointing your child to a trusted grown-up, and flags loudly for you.
- 3
Before your child hears it
Every reply is screened again, and held to a stricter standard than the question. Your child may raise a topic; the AI may never volunteer one.
- 4
Written down, every time
Each decision is recorded — what fired, why, and what the system did. You can audit it, not just trust it.
A sad child is flagged for you, not blocked from the conversation. Blocking distress would teach children that being honest gets them shut out — so NorAI answers warmly and tells you.
Three steps to set up
- 1
Sign in
Google or email. Takes a moment.
- 2
Set up your child
Name, age, interests, language, reading level — and a PIN only you know.
- 3
Hand them the tab
They hold the button and start talking. You watch from the dashboard.
Ready when you are
Setting up the first child takes about two minutes.