SOC 2 Trust Services Criteria
- JWT Authentication — Token-based auth with production secret validation (
src/middleware/auth.ts) - RBAC Permissions — Role-based access control on all sensitive endpoints (
src/middleware/rbac.ts) - Timing-Safe API Key Validation —
crypto.timingSafeEqualprevents timing attacks - Rate Limiting — Enforced in production; fail-fast guard on misconfiguration
- Production Guards — JWT_SECRET validation, demo-mode blocking, rate-limit enforcement at startup
- Health Check Endpoints —
/healthfor load balancer and Kubernetes probes - Circuit Breaker Pattern — Automatic failure isolation for external service calls
- Disaster Recovery — Documented recovery procedures with RTO/RPO targets
- Auto-Scaling — Kubernetes Helm chart with 2-10 replica scaling (Helm + Terraform IaC)
- AI Confidence Scoring — Every mapping includes a 0-100 confidence score with breakdown
- Hallucination Detection — GovernanceService validates AI output against known schemas
- Schema Drift Blocking — Syncs blocked on critical drift; structured
SCHEMA_DRIFT_BLOCKEDresult - Reasoning Traces — DB-persisted step-by-step AI decision audit trail
- DLP/PII Detection — [snapshot]
- Sensitive Data Masking — Shared
maskSensitiveData()utility redacts sensitive field names (password, token, secret, etc.) from all log output - Encrypted Credential Storage — Connector credentials encrypted at rest via ConnectorCredentialService
- GDPR/CCPA Compliance — DLP service designed for EU and California privacy regulations
- Audit Trail Logging — Every AI action logged with user ID, timestamp, and action type
- Data Retention Policies — Configurable retention with 90-day default purge cycle
Compliance Metrics
DB-persisted sessions
Across all providers
Session-scoped event log
Actions passing governance
Regulatory Timeline
EU Artificial Intelligence Act — Full enforcement of high-risk AI system requirements including transparency obligations, conformity assessments, and post-market monitoring.
Colorado AI Act (SB 24-205) — Requires developers and deployers of high-risk AI systems to use reasonable care to prevent algorithmic discrimination.
Built by Auditors
CPA-Grade Audit Trails
Every AI decision is logged with user, timestamp, inputs, outputs, and confidence score — the same rigor Squire applies to client engagements.
SOC 2 Alignment from Day 1
All five Trust Services Criteria mapped to production code. Not a retrofit — compliance was a design requirement from the first commit.
Evidence Export in One Click
Download a structured JSON evidence package covering reasoning traces, AI costs, and audit actions — ready for your SOC 2 auditor.