AI safety promise

For Iseul and the other garden friends

Your child’s garden friend — Iseul the fairy or one of the other friends — can chat using generative AI. These are the rules it always follows.

1. Consent comes first

Free AI chat turns on only after a parent agrees in the Parent Space and confirms with an email code. Without it, your child still plays with the garden friend using set lines — nothing typed is sent anywhere.

2. It says it is AI

Before the first chat the game explains, in words a child understands, that a computer program writes the replies. The chat window carries an AI mark.

3. Risky topics never reach the AI

Self-harm, violence, sexual content, strangers, bullying and personal information are caught before sending. Instead, the game suggests talking to a nearby adult or calling a helpline (Korea: 1388, emergency 112).

4. Personal details are masked

Before anything goes to the AI company, the server trims the chat and hides things that look like phone numbers or addresses. The friend only needs a nickname.

5. Parents can see it all

Read what your child and the garden friend talked about, and what the friend remembers, in the Parent Space.

6. One-tap reports

Any odd reply can be flagged with 🚩. We review reports and block bad replies.

7. Not used for training

Chats are sent to Groq, Inc. (USA) — or Anthropic, PBC (USA) if Groq can’t answer — only to write a reply, under API terms that don’t train on them. We don’t keep raw chat logs on our server.

8. Daily limits

There is a daily cap per child and for the whole service, so the AI can’t be used endlessly. When it’s reached, the friend just gets sleepy.

What AI is not

Replies can be wrong or odd. They are not medical, counselling or legal advice. If your child needs help, please reach out to them yourself, or to a helpline.

Sources: our privacy policy (sections 1, 4, 5) and terms (article 3). Questions: sungwony0906@gmail.com