This is the question every parent, teacher and school asks, and the answer has two halves that people constantly mix up. <b>The model does not learn from your chat. The company may still store your chat.</b> Both are true, and only one of them is about privacy.
Training happened once, months before you opened the app, and then stopped. Typing at it does not update it.
You can test this yourself: tell it something about you, close the app, come back tomorrow and ask. It has no idea.
This is also the annoying half. It will not remember your syllabus, your teacher's preferences or your project tomorrow. You have to explain yourself from scratch every single time.
Your chat is a whiteboard, not a vault — but somebody owns the whiteboard. Most providers store conversations for a period, and some use them to improve future models unless you turn that off.
So the confidentiality question is never "will the model learn it?" It is "who can see the log, and for how long?"
For a school this matters: pupils should not paste other people's personal information into a chat, regardless of what the model does or does not learn.
Assume anything typed into a free public AI tool could be read by a person at that company.
Never paste names, addresses, phone numbers, medical details or anything about another pupil.
If a tool offers a setting to exclude your chats from training, turn it on.
Good news for "will it leak into other people's answers". Not an answer to "is this private".
In our free course you tell the AI something, close it, come back "tomorrow" and ask. The vault is sealed; the whiteboard is wiped.
▶ Play Lemonade Empire free