Baseline Authority for Safe & Interoperable Systems — the open governance standard.
BASIS (Baseline Authority for Safe & Interoperable Systems) is the open specification that defines how AI agents are governed at runtime. It is the foundation that Cognigate implements.
| Component | Purpose |
|---|---|
| Trust Model | 8-tier graduated trust (T0–T7) with 16 behavioral factors across 5 groups |
| Capability Taxonomy | 24 capabilities gated by trust tier |
| Proof Chain | Cryptographic audit trail — dual-hash (SHA-256 + SHA3-256), optional Ed25519 signatures |
| Circuit Breakers | Automatic halt when trust drops below threshold (graduated, with hysteresis) |
| Observation Tiers | 5 levels of agent visibility that cap maximum achievable trust |
| Heartbeat Protocol | Tier-scaled liveness checks with deadman switch |
| CAR | Canonical Agent Record — unique, verifiable identity per agent |
Trust scores are computed from 16 behavioral factors organized into 5 groups:
| Group | Factors |
|---|---|
| Identity | Registration completeness, credential validity, identity continuity |
| Behavior | Action compliance, boundary respect, escalation appropriateness |
| Performance | Task completion rate, response quality, resource efficiency |
| Security | Input sanitization, output validation, anomaly frequency |
| Governance | Policy adherence, audit trail completeness, recovery behavior, operational context |
BASIS sets the rules.
CAR identifies the agent.
INTENT figures out the goal.
ENFORCE stops the bad stuff.
PROOF shows the receipts.