Security architecture
Moirai keeps clinical AI governance records isolated, encrypted, exportable, and verifiable. The security story is not a badge wall. It is a chain of custody a buyer can inspect.
Control plane
Active layer
RLS enforced
scope:practice
Application permission checks derive the practice from the authenticated profile, then database RLS enforces the same boundary at table level.
Security pages fail when they only say the safe words. This surface states what exists, what is prohibited, what is independently certified by vendors, and what Moirai has not yet completed.
Authenticated reads and writes are scoped through Supabase session cookies, server-side permission checks, and Row Level Security policies.
Full exports and entity exports require audit permissions, are rate-limited, and append reviewer-visible activity events.
Moirai is designed for governance metadata. PHI is prohibited and evidence uploads include conservative review flags.
Moirai does not yet hold SOC 2 or ISO 27001 certification. Current posture, limits, and planned assurance work are published in the Evidence Room.
Control map
The architecture reads from identity to public verification. Each layer exists because clinical governance records are sensitive even when they contain no patient data.
01
Supabase Auth issues httpOnly cookies. Protected routes refresh sessions through middleware before page or API access.
02
Application permissions derive the practice from the authenticated profile. Database RLS remains the second lock, not the only lock.
03
Uploads are validated for type, size, integrity hash, and PHI-risk signals before they can contribute to a Follow-Up Evidence Pack.
04
The verifier exposes only PHI-excluding metadata for hashes, report IDs, and evidence events.
Data boundary
Moirai is designed for governance metadata: AI tools, owners, approvals, evidence references, policies, reviews, and exports. Patient-identifiable health information is outside product policy.
Sub-processors
Due diligence should not require email archaeology. The review path links the DPA, security limits, and public verifier.
| Provider | Purpose | Location | Assurance |
|---|---|---|---|
| Database, authentication, storage | Sydney, AU (ap-southeast-2) | SOC 2 Type IIHIPAA | |
| Hosting, edge, serverless functions | Sydney PoP, global edge | SOC 2 Type IIISO 27001 | |
| Payments and billing | US / EU | PCI DSS Level 1SOC 2 | |
Sentry | Error monitoring, performance telemetry | US | SOC 2 Type IIGDPR |
PostHog | Product analytics, feature flags | EU | SOC 2 Type IIGDPR |
Resend | Transactional email delivery | US | SOC 2 Type II |
Loops | Lifecycle email and user communications | US | DPA listed |
| Governance content generation (no patient data sent) | US | SOC 2 Type II |
External references
These are evidence mapping references and certification roadmap items. They are not claims of regulator approval, clinical validation, or third-party certification.
Found a vulnerability? Email the security team directly. We take reports seriously and respond through the incident-response path.
Due diligence
Open the Evidence Room for current posture, certification limits, sub-processors, documents, DPA, and public verification.