AI Memory and Personalisation: Helpful or Creepy?

Many assistants can now remember details about you across conversations. That can make them more useful, and raises privacy questions.

• Saved preferences: your tone, role, projects and dietary needs.
• Past-conversation recall: referring to earlier chats.
• Custom instructions: standing rules you set.
• Project knowledge: documents and notes attached to a workspace.

It usually works by storing notes or summaries and inserting them into new conversations, not by the model "learning" in real time.

• Less repetition: no need to re-explain your context.
• Better tailored answers.
• Continuity on long projects.

• Sensitive details stored longer than you expect.
• Wrong or outdated memories shaping answers.
• Over-personalisation: a bubble that reinforces your views.
• Data use: is memory used for training, shared or accessible to others?
• Shared devices and accounts revealing personal information.

1. Find the memory settings and review what's stored.
2. Edit or delete anything wrong or sensitive.
3. Turn it off or use temporary/incognito chats for private topics.
4. Avoid storing health, financial or legal secrets unless you're comfortable with the provider's terms.
5. Use separate projects or accounts for work and personal use.
6. Check retention and deletion policies.

Let it remember preferences and context, not secrets.

Memory makes AI feel more like an assistant who knows you. Keep it tidy, check what's stored and decide what you're happy for it to know.