Playground safety and privacy
How Playground conversations are checked for safety, what is stored, and how parents can review them.
Who can use the Playground
The AI Playground is limited to learners aged 13 and over. Accounts registered as under 13 cannot open it, and it does not appear in their navigation.
Every message is checked
Playground conversations are checked for safety in both directions: what a learner sends is reviewed before it reaches the AI model, and the reply is reviewed before it appears on screen. Anything that fails those checks is not shown.
Ren and Mira are also instructed never to ask for personal information, never to give medical, legal, or financial advice, and to decline romantic or sexual topics.
If a learner seems to be in distress
If a conversation suggests self-harm or a crisis, the character encourages the learner to talk to a trusted adult right away and points to the 988 Suicide & Crisis Lifeline in the US. It does not press for details. Conversations flagged this way are queued for review by our staff.
Automated checks are not a substitute for adult attention. If you are worried about a young person, please talk to them directly and contact a local crisis service.
What is stored
We store the messages a learner sends in the Playground and the replies they receive, so that chats can be reopened later. Learners can see only their own chats, and search never reaches another learner’s conversations.
Playground chats are deleted automatically 12 months after their most recent message. Chats a learner deletes are removed from view immediately and erased permanently about 30 days later.
How parents can review Playground activity
Playground conversations are included in the account data export, so a parent can read what their child has created. See Requesting data review or deletion for how to export data, delete an account, or send us a request directly.
AI providers
Playground messages are processed by our AI provider in order to generate a reply. We do not sell learner data, do not use it for advertising, and do not allow it to be used to train external models. The providers we rely on and what they receive are listed in our Privacy Policy.