Security
Huddles contain the sensitive part of a company's thinking: the disagreement before the decision. This is how we handle it, in plain terms.
Access
Every huddle, post, call, reading and decision belongs to one workspace. Access is enforced in the database on each row, not only in the interface, so a request can never read another workspace's data even if it asks for it directly.
Inside a workspace, people see the huddles of the usergroups they are a seat in. Invitees see the huddle they were called into.
Authentication
Sign-in runs over Google or email with a hashed password; sessions are bearer tokens validated on the server for every request that touches data. API keys are workspace-scoped and revocable.
Encryption
All traffic uses TLS. Data is encrypted at rest, and backups are encrypted with it.
AI
- The AI does two things only: tag a post, and produce one stored reading per huddle.
- The reading is stored, not recomputed on every view, so your content leaves our systems as rarely as possible.
- Your content is not used to train models.
- Every AI tag can be corrected by a human, and the correction wins.
What should never happen
- A member of one workspace reading another workspace's huddles or decisions.
- An invitee seeing huddles they were never called into.
- A decision record silently changing: reopening keeps the earlier version.
Subprocessors
Hosting and database (EU region), the AI provider used for tagging and readings, and email delivery for notifications. We will tell you before adding one that touches huddle content.
Reporting something
Found a weakness? Tell us before telling anyone else and we will respond within two business days.