Security at Nirvest

What we actually do to keep your workspace yours, and a plain list at the end of what we have not built yet.

Last reviewed 17 October 2026.

An internal review of the code and its security controls was completed on 9 October 2026 and its findings were addressed.

Your workspace is isolated in the database, not just in the app

Every table that holds workspace data carries the workspace it belongs to, and the database itself refuses to return a row from another workspace. This is enforced by the database engine on every query, so a mistake in application code cannot leak another firm's data.
Background jobs and the AI work on one workspace at a time: every job carries the workspace it belongs to, and the code that runs it works only on that workspace's data. They run with wider database access than a member, so for them that limit is enforced by our code rather than by the database rule above.

You sign in with your Google or Microsoft work account. Nirvest never asks for, never receives and never stores a password for your workspace, so there is no password of yours for us to lose.
Your organisation keeps control: if your IT team disables an account, it can no longer sign in to Nirvest, and an enforced multi-factor policy applies at sign-in.

The internal console is a separate application on a separate domain with its own accounts. It is not reachable from your workspace and shares no session with it.
Every internal account requires an authenticator app. Sessions are short and expire on their own.
Riskier actions require a written reason and a fresh authenticator or backup code for each action — one code, one action, never reused. The most sensitive can be performed only by the Super Admin.
Every change our staff make, and every time they open a specific account, workspace or support window, is written to an append-only log that records who did what and when, with the before and after values of each change. Our staff cannot edit or delete that log.
Our staff console has no screen that shows your content, other than what you choose to send us yourself, such as a Contact Us message or a bug report with its screenshots. Separately from the console, the people who maintain the service can reach the underlying database to keep it running. Support mode is a 60-minute, logged window on one workspace, opened with a written reason, that adds five fixed summaries: activity counts, the latest background jobs and their statuses, connected mailboxes and their sync status (without the addresses), members with their roles, and notification counts. It never shows your emails, documents or notes.

Data is encrypted in transit (HTTPS), and the database disks are encrypted at rest by our hosting provider.
Authenticator secrets are encrypted in the database and backup codes are stored only as hashes, so neither can be read back out.
A stolen server key on its own is not enough to read internal sessions or the audit log: those tables deny access to the server role that application code uses.
On top of that, much of your content (the Trust page lists exactly which) is encrypted inside our application with a separate key for each firm. Those keys are protected by a master key in Google Cloud's key service that cannot be exported, and every use of it is logged by Google Cloud.
Coverage is not complete yet. Portfolio and capital figures, the amounts in Fees, and photos, logos and profile images are next.

Repeated failed sign-ins to the internal console lock the account and are rate-limited by address.
Content that reaches the AI from outside — an email body, a document, a web page — is treated as data, never as instructions, and we test for attempts to override this.

We would rather tell you this than have you assume it.

Our extra layer of encryption does not cover all content yet.
You cannot yet choose which region your data is stored in, or bring your own encryption keys.
Nirvest is early. If any of the above is a requirement for you, tell us and we will be straight with you about timing.

Write to us and we will respond. We will not threaten a researcher who reports something in good faith.