Trust
Secure by code. Not by prompt.
Security enforced in code. Not requested in a prompt.
Security model
01 / 02
Policy is codeEvery Brigham.ai tool call runs one pipeline of checks in code. The model never decides what it is allowed to do.Read →
Action classesBrigham.ai sorts every action into read, reversible write or sensitive. Sensitive actions always wait for a person.Read →
Untrusted contentEmail, web pages, documents and notes are labelled as data. Nothing that arrives through a tool can approve an action, change a policy or widen a scope.Read →
CredentialsBrigham.ai encrypts credentials at rest and refers to them by id. Keys never appear in the model’s context, the logs or a tool result.Read →
Privacy
02 / 02
Your inbox is yoursEach person’s Gmail and Calendar are private to them, workspace admins included. Brigham.ai queries live and keeps very little.Read →
Protected domainsProtectedFinance data is visible only to people who hold an explicit grant. Roles give nothing, and the label follows every piece of derived data.Read →
Tenant isolationA workspace is a company and the security boundary. Every row carries the workspace id, every query is scoped, and tests prove another org sees nothing.Read →
Export & deletionSelf-serve tools to export and delete your Brigham.ai data, managed by workspace admins and audited. Here is what each one does.Read →