Safety commitments
Tala is used by children aged 3 to 12. This page is our public statement of how we protect them, written for parents, teachers and anyone who wants to check our work.
For the details of what data we collect and why, read the privacy notice. Kids can read a version written for them.
Last updated 3 November 2026
Parents own the account
- Only an adult can create a Tala account. Kids cannot sign up on their own.
- When you sign up you give consent, as the child’s parent or guardian, for your children to use Tala, and you agree to our terms and privacy notice.
- Kids learn in kid mode on your device. A 4-digit parent PIN, which we store only as a salted hash, is the only way back to settings, billing and chat transcripts.
We collect as little about kids as we can
For each child we store a nickname, a Kislap colour and an age band (Little 3–5, Explorer 6–8 or Thinker 9–12), plus their lesson progress, the Dream Paths they follow and their Kislap chats. We never ask a child for an email address, phone number, photo, school, birthday or real full name.
Kislap is an AI, and says so
AI in Tala works mostly in the background: it chooses what to review and what comes next, and fills vetted lesson templates with familiar examples. It never writes facts freely.
- Under 6, there is no open chat. Little learners hear stories read aloud and answer a few spoken questions inside a lesson. They cannot type or talk to Kislap freely.
- From age 6, kids can ask Kislap for help. The chat screen always tells them that Kislap is an AI helper, not a person.
- Kislap is a learning helper, not a friend or companion. It does not roleplay and steers off-topic chats back to learning.
- It teaches the Socratic way: for homework-style questions it gives a hint or a guiding question, never the final answer.
- Messages are limited to 400 characters, and each child has a daily message cap (5 on the free plan, 40 on Plus).
Filters run before the AI sees anything
- Personal information. Our server checks every message a child sends for phone numbers, email addresses, street addresses, school names and full names. A message that contains them is held back and never sent to the AI. Kislap explains, kindly, why we keep that information private.
- Distress and sensitive topics. Messages about self-harm, abuse and similar topics get a fixed, safe reply written by us, not by the AI. It tells the child to talk to a parent or trusted adult and gives the National Center for Mental Health crisis line, 1553. The message is flagged for the parent.
- A child-safety system prompt guides every reply Kislap gives.
Parents see every chat
Every message between your child and Kislap is saved to your account. You can read the full transcript for each child in the parent area, with flagged messages highlighted.
Designed for calm, not for more screen time
- No ads, no autoplay, no endless feed, no in-lesson purchases, and no prices ever shown to kids.
- Every session ends on purpose with an “All done! Time to go play” screen and something to try away from the screen.
- No public leaderboards and no chatting with other users. Siblings can see a private family board.
- Rewards are tied to skills, never random, and streaks can never be bought.
- You set a daily limit of 15, 30 or 45 minutes, and Tala keeps it.
- Music is off by default, and motion stays gentle. Reduced-motion settings are respected.
- No guilt messages, countdown timers or requests for kids to ask you to pay.
Which AI we use
Kislap’s replies are generated by Claude, a model made by Anthropic, reached through Vercel AI Gateway. We follow Anthropic’s guidelines for organizations whose products are used by minors: we tell kids they are talking to an AI, we moderate and monitor conversations, we design for their age, and we publish this statement. If we ever find child sexual abuse material, we report it to the proper authorities.
The laws and guidance we follow
- We handle personal data under the Philippine Data Privacy Act of 2012 (Republic Act 10173), which treats anyone under 18 as a child whose parent gives consent.
- We publish a privacy notice that kids can read, following the National Privacy Commission’s guidance on child-oriented transparency (NPC Advisory No. 2024-03).
- We designed Tala in line with the principles of the US Children’s Online Privacy Protection Act (COPPA): parental consent, minimal data from children and parent access to their child’s data. We are not certified under any COPPA safe-harbor program.
- We never sell personal data and never use it for advertising.
If a child is in danger
If someone is in immediate danger, call your local emergency number. In the Philippines that is 911, and the NCMH crisis line for a mental health crisis is 1553. Tala is a learning app and cannot respond to emergencies.
Report a concern
If you see something in Tala that worries you, such as a Kislap reply that is unsafe, wrong in a harmful way or not right for a child, please tell us. Email help@tala.kids with:
- what happened and roughly when,
- the lesson or chat it happened in, if you know, and
- the email address on your account, if you have one.
Please don’t include your child’s full name or other personal details. A real person reads every report.