AI glossary
Agent memory
Information an AI assistant or agent keeps between conversations, such as your preferences or facts about a project, so it does not start from zero each time.
Models themselves do not remember past chats; each request starts fresh. Memory is added around them: the system saves notes (in files, a database or a vector database) and puts the relevant ones back into the context when needed.
Memory makes assistants more useful but raises privacy questions. Check what your assistant remembers, where it is stored and how to delete it.
Example: You tell your assistant that you write in British English and prefer short answers. Weeks later, in another conversation, it already answers that way. The model does not remember it; the app saved a note and adds it to the context.
In practice
- Check in the settings what your assistant has saved and delete anything you do not want it to use.
- Do not share sensitive details “so it remembers”: they are stored like any other data in your account.
- If you build an agent, keep only what will be useful later (decisions, preferences, task status), not whole conversations.
There is more on what happens to your data in our guide to hallucinations and privacy.
