Version 1.0
Protection charter
This is not a privacy policy written by lawyers to protect us. It is the list of what we do, what we will never do, and what you can verify yourself.
1. Our commitments on artificial intelligence
No answer reaches a child without being reviewed
A first model drafts. A second model, from a different provider, reviews before display and can refuse. If that second model is unavailable, we apply a strictly deterministic mode rather than let an unchecked text through.
The AI always declares itself as such
Piko tells the child it is a program, from the very first message. It refuses to be called "friend", never says it loves the child, and never builds emotional dependency. A six-year-old cannot make that distinction alone: it is on us to keep reminding them.
The AI admits its mistakes and encourages doubt
Piko flags that it can be wrong on every important fact, and congratulates the child who catches it out. We believe learning to doubt a machine is the most useful skill we can pass on today.
The AI does not do the child's work
In the "Create" World, Piko never produces the finished work. It asks a question, offers at most two directions, and hands back control. The child stays the author.
The AI does not try to keep the child
No notifications, no re-engagement, no streaks to maintain, no infinite scroll. The time budget is enforced server-side, and Piko itself closes by telling the child to go play outside.
2. Personal data
- 1
The child has no account. No email, no password, no identifier. A first name — or a nickname — and an age, under their parent's account. There is nothing to steal.
- 2
Personal data is erased before anything is sent. If a child writes a phone number, an address or an email, the filter removes it on our server — before the message goes to any AI provider.
- 3
No third party observes the child. No CDN, no remote font, no analytics, no advertising pixel, no iframe. The browser security policy technically forbids any outbound request from our pages.
- 4
No resale, no training. Conversations are never sold, never transferred, never used to train a model.
- 5
You take it away or erase it, whenever you want. Full export in one click from the parent space, complete deletion of the account and history in another. Without writing to us, without delay, without conditions.
3. Transparency towards the child
The parent sees every exchange. We refuse to let that happen behind the child's back: it would teach them exactly what we are fighting — the idea that a system can watch you without telling you.
That is why Piko's rule number 3, announced by Piko itself in the very first session, is: "Your parents can read everything we say. Nothing is hidden." And under every answer, the child sees two labels: who wrote it, and who reviewed it.
4. What Cogency-Kids is not
It is not…
- A substitute for a parent, a teacher or a health professional
- A mental health support service
- A source of truth — Piko gets things wrong, and says so
- A way to leave a child alone in front of a screen
It is…
- A framed companion for curiosity, under adult responsibility
- A tool for AI literacy, from the earliest age
- A space where a machine's mistake becomes a lesson
- A starting point towards the real world, never a destination
5. In case of danger
If a child mentions danger or distress, Piko switches immediately to support mode: it acknowledges, says it is not the child's fault, and directs them to a trusted grown-up — without asking investigative questions and without giving medical advice. The parent receives a critical alert in their space.
Cogency-Kids replaces no emergency service. In case of immediate danger, contact the relevant authorities in your country.
Questions about this charter?
Write to us. A charter you cannot argue with is not a charter.
Charter version 1.0 · Cogency-Kids, a vertical of the Cogency platform · Synaptics Solutions & Technologies Ltd, N'Djamena, Chad.