Skip to content

Security

Each organization kept apart.

How Global Sales AI protects your data and your contacts' choices: what the product does, described plainly.

(01)

Isolation in the database

Every table that holds an organization's data is protected by PostgreSQL row-level security. Organization data is read and written through a database role that sees only the rows of the signed-in organization; without that context it sees nothing.

An automated test fails if a table with organization data lacks these policies.

(02)

Roles and permissions

Five roles, from owner to viewer, decide who can read, edit, approve and manage. Permissions are checked on every action, and again for sensitive ones such as approvals, closing deals, roles and API keys.

Discounts, proposals, contracts and payment links need a second person: whoever asked for them cannot approve them, unless they are the owner.

(03)

Sign-in

Passwords must have at least 12 characters and are stored as scrypt hashes. After five failed attempts an account is locked for 15 minutes, and sign-in and registration are rate-limited.

Each person can turn on two-step verification with an authenticator app (TOTP) and gets ten single-use recovery codes. Session cookies are HttpOnly, and only a hash of the session token is stored.

(04)

Credentials and records

Keys and passwords of connected services are encrypted with AES-256-GCM and bound to your organization and to the service. The interface shows only a fingerprint, never the secret.

Every change of state is written to an audit log with who made it and when. The application can add to the audit and security logs, but it cannot change or delete their entries.

(05)

Your contacts' choices

Every message tells its recipient how to unsubscribe, and e-mail programs get a one-click unsubscribe header. Unsubscribes and withdrawn consent go to the suppression list, which is checked again at the moment of sending.

When a contact is erased, only a keyed hash and a masked address stay on the suppression list, so that the person is not contacted again.

(06)

AI within limits

Text from outside, such as lead records, imported files, inbound e-mails and web pages, is treated as data, never as instructions. It is checked for injected instructions, and every model answer is validated against a schema before it is used.

With the platform's AI, spending has a monthly budget per plan. With your own Anthropic key (Enterprise) the plan budget does not apply, but each agent can still have its own monthly budget. Agents have no tools that act outside the platform on their own: in the chat, steps that send data out wait for a person's approval, and every external message goes through the approval queue.

(07)

Your data, your exit

The owner can download the organization's data at any time as one ZIP file, with a JSON and a CSV file per table (up to 50,000 rows per table), in every plan and during the demo. Passwords, secrets and credentials are never part of it.

Requests from people whose data you hold (access, erasure, rectification, objection) are recorded with their due date. The audit log can be exported as CSV.

This page describes what the software does. It is not a certification and not legal advice.

See the controls for yourself.

Create an account to find these settings in the app: members and roles, two-step verification, the audit log. The plans are on the pricing page.