Security and deployment.

Where does it run?

One of two ways. Managed: your own isolated tenant on the Torus cloud, and data sent to the model is dropped after each request. Sovereign: inside your own cloud, VPC or on-premises servers, with the model inside, and nothing leaves your network. It is the same product either way.

Is our data used to train models?

No. Your data never enters model training or a shared embedding space. Frontier models run under zero-retention terms. Your data improves only your firm's own processes, inside your tenant.

Whose models, and can we keep them off-platform?

Your choice. Frontier models by default, or Torus Sovereign: open-weights models, self-hosted and tuned to drive the agent, entirely on your infrastructure.

Who inside our firm can see what?

Role-based access, with row-level security at the database. Queries are scoped to your firm, and what each person sees is limited by role.

Is Torus SOC 2 compliant?

SOC 2 Type 1: complete. SOC 2 Type 2: in progress.

Who approves the work?

Your engineer. Your engineers approve each sheet before final issue, and your engineer is credited as the preparer. Torus drafts; the engineer decides. The published packages are at issued for review, before the engineer of record's review and seal, so their signature and seal fields are blank.

Where does each value come from?

Every value is traced to its source: the sheet, revision and cell it came from. Each value is checked against every other sheet that carries it before the package is issued.

How do we report an incident?

Customers report incidents, concerns and service issues to support@usetorus.com, which is also shown inside the Torus application. It is a monitored mailbox: each report is tracked, triaged by severity and followed to resolution.

Where do security questions go?

Write to security@usetorus.com with security questions or a review request.

Get the demo packageRead the FAQ