Safety
Five layers, before a child ever sees a word of it
This is a plain-language account of what the code actually does, not a policy statement written separately from the product. For the legal version, see Privacy and Children's Privacy.
- 1
Age-band content policy
Every generation and every tutor turn is governed by a policy for the learner’s specific age band (7–9, 10–12, 13–14, or adult) — not one generic filter for everyone.
- 2
Grounded sourcing
Courses are built from real, retrieved material. A lesson must carry at least one citation that maps to a source the research step actually returned — content without a valid citation fails validation and is never shown.
- 3
Safety classifier
A dedicated safety model checks every generated lesson and both directions of every tutor turn against the learner’s age band. Unsafe input from a child never reaches the generation model; a flagged reply is regenerated once, then replaced with a safe canned response.
- 4
Parental approval
A child’s course request either follows a category rule the parent already set, or waits — with the finished content attached — in the parent’s approval queue. A denied category is blocked at request time, before anything is generated.
- 5
Answer isolation
The correct quiz answer is never sent to the client ahead of time. The tutor is given lessons through the same filter, so it cannot leak an answer even if asked directly.
Reporting
A report button that actually does something
Every generated lesson and every tutor message carries a report control. Reporting hides the content immediately on the reporter's device and queues it for review — it does not wait for a moderator to act before protecting the person who reported it.
Privacy
Built to send the model almost nothing
Only a first name + age band
reach the AI model, ever — through one point in the code, so it can’t leak by accident.
Hashed analytics identifiers
no raw child identifiers ever reach the analytics store.
No ad targeting, any age
and no data sale, on a child’s data or an adult’s.
Full legal detail: Privacy Policy · Children's Privacy
Compliance
Aligned to COPPA, India's DPDPA, and the Australian Privacy Act
Every child's account requires verifiable parental consent, applying DPDPA's stricter under-18 standard everywhere we operate, not only in India. A named Grievance Officer is available for India-specific requests — see Children's Privacy.