Security
Enterprise security. AI governance.
Every AI request passes through 7 security layers. Every user action logged. Every data access traceable.
AI Security Pipeline
Seven layers on every request.
1. Team Guard
→
2. Kill Switches
→
3. Rate Limit
→
4. Budget
→
5. Scope
→
6. LLM
→
7. Audit
Kill Switches
3 levels — global, division, team. Instant shutdown. Zero cache delay.
Scope Registry
Declarative data boundaries. Safety agents can't query Finance models.
Rate Limiting
Per-user, per-team RPM via Redis. 60 RPM default, configurable.
Token Budgets
Daily budget per team. Warning 80%. Hard stop 100%.
API Key Mgmt
Per-team encrypted keys. Rotation, model allowlists, bulk update.
Security Monitor
Validates access every request. Violations logged with full context.
Platform Security
Enterprise authentication and data protection.
SSO
Enterprise identity management
MFA
TOTP for admin/exec roles
Audit Trail
Every mutation logged
Encryption
Bcrypt 12-round, TLS
RBAC
12 roles, dept permissions
Rate Limiting
Auth: 5 req/60s, lockout 5 fails
GDPR
Export, erasure, anonymization
CCPA
Do-not-sell flag support