Trust & security
The one-page answer for your security team, in plain language. If a question isn't answered here, ask — we'll send the document, not a sales line.
Data handling
Your data is isolated in a per-tenant environment, encrypted in transit and at rest, and never used to train any model. There is no shared model to leak into — the guarantee is structural, not just contractual. Cipher suites and key handling are specified in the security package.
Identity & access
agntmind connects through scoped OAuth and enterprise connectors — not stored passwords or screen-scraped sessions. Every action runs under least-privilege, role-based permissions, scoped to what that user may touch, and integrates with your identity provider so MFA is preserved. Ask for the connector and SSO coverage list.
Human approval
These action classes always require a human yes and can't be turned off: moving money, external communications, deleting records and executing contracts. You set the rest — and you can stop any run mid-task.
Audit & logging
Every action is logged immutably with attribution (who, what, when, why, approved-by), and exportable to your SIEM.
Certifications
Honest status — no badge until the report exists.
Subprocessors
Every model and infrastructure provider in the data path is listed in the security package and kept current as models change. Subscribe to change notices.
Ownership & exit
Your Brain and data are yours. Export on request, and full deletion after termination on the schedule set in your DPA. You are never held by the switching cost.
Incident response
Breach notification within the window committed in your DPA, with a named contact and a written post-incident report.
Your security team gets documents, not sales lines.