Security
AI that acts needs more than talk about encryption. Here’s where I’m checked, step by step, with a record no one can quietly edit.
One enforcement path, six checks — power is re-checked at every layer it passes through.
Gate 1 · Planning
A step that needs your OK can't slip into a plan. It turns into a question for you instead.
Gate 2 · Handing out work
Work only goes to people and agents allowed to do that kind of work. Not allowed, not assigned.
Gate 3 · The pass
Each run's pass is tied to one company, one actor, one mission, a set of rights and an end time. Agents get the notes for the job, never database access.
Gate 4 · Each call
Each call is checked again on our server against today's rules and rights. The pass is a hint; the server decides.
Gate 5 · The check
No one can skip a required check, people included. An override is recorded as a decision.
Gate 6 · Release
Checked work still waits for someone allowed to release it. Verify holds work back; your limits and your people let it through.
Each company kept apart
Each company's data is walled off in the database itself. We attack it with our own tests before each release. That's how we work, not a certification.
Sign-in and roles
Sign in with Google or your email on every plan; enterprise SAML/SCIM comes after launch. Access by role, and you choose who can see which facts.
Each task sealed off
By kind of task: data (a pass for one job), code (its own copy of the code), browser (approved sites only), messages (drafts first), finance (fixed limits and a person's OK).
AI providers
They get only what a job needs, keep no lasting access, and never train on your data without your clear consent.
The record
Each action goes into a chained log that shows any edit. Anything that matters can answer what it used, what it did, what it cost and who said yes.
Your data, start to finish
Export it anytime. Ask us, and we delete it. The DPA sets out the details.
Uptime
Uptime is published on status.brainis.ai. It appears here once the status feed is wired — we would rather show nothing than a number we did not measure.
Security questions: support@brainis.ai · past notices are on the Trust page.