A bot cannot read direct messages between other people. It sees conversations it is part of.

What if someone leaves the company?

Scoping is checked at retrieval against current membership, so access changes take effect on the next question rather than requiring a re-index.

Can it post back into Slack?

Yes, and posting is a write, so it asks for approval first unless an admin has set that specific action to always allow.

Can an AI read our Slack without seeing private channels?

Yes. A bot only sees the channels it has been invited to, and a well-built layer additionally checks who is asking before returning anything.

Slack's own model already limits a bot to channels it has joined.

The harder problem is that two people asking the same question should get different answers.

That requires resolving the asker to a real identity, not just authenticating the app.

Two different questions

The first is what the integration can read at all, which Slack governs: a bot sees public channels plus any private channel it has been explicitly invited to.

The second is subtler and more often got wrong: once that content is indexed, does the system remember who is allowed to see which part of it? If it does not, anyone who can ask a question can reach everything that was ever ingested.

Why per-person scoping is the hard part

Answering as an application is easy. Answering as a specific person requires resolving that person across systems, so that a Slack user maps to the same human as a dashboard login.

Indexbrain resolves Slack users to accounts by email and applies the same scope checks it applies on the web, so an answer in Slack respects personal, department and room scoping rather than returning whatever was indexed.

What is kept

Long-term retention of raw messages is its own risk. Indexbrain purges source content on a rolling 90-day cycle; what persists is the extracted, sourced knowledge rather than the messages themselves.

Try it free See it work