Browse documentation

Your privacy

A plain-language summary of what Cognos can and can't see — and the controls that keep it that way.

3 min read Updated 22 July 2026

1What we can't see

The content of your saved conversations, attachments, bookmarks and memory is kept in a form we can't read. It's unlocked only on your device, with your Account Key.

2What we can see

  • Metadata needed to run the service — when a message was saved, and what a request cost.
  • The AI provider that answered each message (recorded on the answer).

The AI provider does see your message while it answers — that's how any AI assistant works. Privacy tiers control where that happens.

3Check any answer

The Security button on a conversation shows, for each answer, which model and provider handled it and where — so you can verify it matches your settings.

A modal comparing what the account holder sees with what the model saw.
You always see your real values; the model only saw placeholders.

4Your controls

Privacy tiersLimit where messages may be handled.RedactionHide sensitive details before sending.Auto-deleteMake content delete itself.MemoryDecide what the AI remembers — privately, or shared in a project.

Keep reading

Privacy tiersYour privacy tier sets the furthest your messages may travel to be answered. It changes which models you can use — and nothing else.RedactionCognos can spot sensitive details in your message — an IBAN, an email, a card number — and hide them before the AI ever sees them.MemoryMemory lets the AI remember facts and preferences you choose — across your chats, or within a project.