Security overview
Public security posture for Surfaice: permissions, isolation, subprocessors, and compliance status.
Warning
Canonical source: Compliance claims (SOC 2 status, subprocessors, training, contractual commitments) live on surfaice.pro/security. This page summarizes product behavior for doc readers and must not outclaim that page.
This page summarizes security topics for product users and reviewers. The live, questionnaire-oriented write-up lives on the marketing site and is the preferred link for IT reviews.
Hugo is Surfaice’s in-app AI agent. When other pages say “agent requests” or “agent runs,” they mean work Hugo performed in a customer workspace.
Enterprise security summary, subprocessors, compliance status, and NDA package request.
Highlights
- Permissions inherited from source systems; Surfaice does not expand entitlements — see Permissions
- Per-user mailbox connections — no tenant-wide mailbox read
- Isolated customer environments on Google Cloud (US regions)
- No customer data used to train models (contractual)
- Audit trail of agent requests — who asked, what tools ran, what was produced
- SOC 2 Type I — in progress via Vanta, expected Q3/Q4 2026; see surfaice.pro/security
In-product trust docs
These pages complement the marketing security summary:
- Trust FAQ — procurement and compliance questions
- Audit trail — what is logged and who can access it
- Agent limitations — honest limits on memory, corpus coverage, and accuracy
- Data handling — persistence model and subprocessors
Documents under NDA
Request the Security Guide, Application Architecture Diagram, and AI Governance Policy via security@surfaice.pro.
Vulnerability disclosure
Report security issues under our vulnerability disclosure policy.