Trust

Security & compliance

This page documents controls that are actually implemented, and is explicit about what is not. No certification is claimed that has not been independently audited.

Implemented

Authentication

scrypt password hashing, seven-day sessions stored as hashes, optional TOTP, lockout after repeated failures.

Implemented

Authorization

Four roles with server-side permission checks on every mutation; owner-only admin surface.

Implemented

Tenant isolation

Every tenant query is scoped by workspace id. Cross-tenant access attempts are refused and logged as critical.

Implemented

Secrets

Integration credentials encrypted with AES-256-GCM; API keys stored as SHA-256 hashes and shown once.

Implemented

Input handling

Parameterised SQL everywhere, allow-listed outbound HTTP, SSRF blocking, strict security headers and CSP.

Implemented

AI safety

Tool allow-lists, approval checkpoints, prompt-injection defences, redaction and full execution audit.

Not claimed

No SOC 2 Type II, ISO 27001, HIPAA or PCI DSS attestation is claimed. Card data never enters this application — payments are hosted by Stripe. Regulated workloads (healthcare, financial advice, legal decisions) require a signed agreement and a completed compliance review before production use.

Security & Compliance

This page documents the controls actually implemented in the platform. It deliberately does not claim certifications that have not been independently audited.

Implemented controls

  • Passwords hashed with scrypt; optional TOTP multi-factor authentication.
  • Session cookies are HttpOnly, SameSite=Lax, signed and stored only as hashes.
  • Role-based access control (owner, admin, member, viewer) enforced server-side.
  • Tenant isolation: every tenant query is scoped by workspace id; cross-tenant access is rejected and logged.
  • Integration credentials encrypted at rest with AES-256-GCM.
  • API keys stored as hashes and shown once.
  • Rate limiting on authentication, public forms and the API.
  • Audit logging for security-relevant actions, plus a security event feed.
  • Strict security headers and a restrictive Content-Security-Policy.
  • AI safety: tool allow-lists, per-agent scopes, human approval checkpoints, prompt-injection defences, output validation and redaction.
  • Data export and deletion tooling for customers.

Not yet certified

No SOC 2, ISO 27001, HIPAA or PCI attestation is claimed. Card data never touches this application: payments are handled by Stripe. Regulated workloads (e.g. healthcare) require a signed agreement and a completed compliance review before production use.

Security and compliance · Sonai