AI Actions & Approvals

3 Action Levels

Automatic actions

Read-only. No risk. - Viewing data - Scanning vulnerabilities - Analyzing threats

Security actions

Instant with confirmation badge. - Blocking IPs - Adding honeypots - Enabling TLS - Setting rate limits

Destructive actions

Need final approval. - Deleting rules - Deleting agents - Deleting assets

Why this design?

- Read actions can't hurt anything - Write actions are logged but reversible - Delete actions are permanent — need explicit approval

Audit Trail

Every action is logged: - Who did it (you or AI) - What was done - When - Result

View in Compliance → Audit Log.