System Governance
QUORUM enforces a rigorous governance model that ensures every decision is traceable, every rule is validated, and every override is authorized within a strictly defined hierarchy.
Rule Lifecycle
Create & Stage
Rules are defined in a declarative schema and staged in a non-execution environment.
Shadow Evaluation
Every candidate rule runs in shadow mode against historical event streams before promotion — quantifying precision impact, false positive rate, and coverage before any live traffic is affected.
FROST Threshold Approval
Deployment requires a (t,n) cryptographic quorum from designated compliance officers via FROST(Ed25519). No single officer can unilaterally promote a rule. The approval record is an unforgeable threshold signature stored in the audit ledger.
Live Deployment
Atomic promotion to execution nodes with sub-second propagation. The promotion event is KZG-committed to the audit ledger.
Approval Thresholds
Network parameters, IP reputation sensitivity, and caching TTLs. Single-engineer approval with full audit trail.
Risk score weights, block/allow thresholds, and model sensitivity. Dual-compliance FROST co-signature required.
Global bypasses, network-wide isolation, and consensus resets. Institutional executive sign-off via full FROST threshold quorum.
Override Logic
Strict hierarchy ensures that manual overrides are recorded with full principal attribution, automatic expiration timers, and a counterfactual explanation certificate generated at override time.
Conflict Resolution
The Adaptive Rule Engine resolves rule-vs-model conflicts according to pre-configured precedence policies (Strict, Balanced, or Technical). All conflict resolutions are logged with their governing policy version.
Change Traceability
Every administrative action is hashed and KZG-committed in the append-only audit ledger. O(1) inclusion proofs allow any party to verify any historical change without traversing the full chain.
Counterfactual Certificates & Regulatory Explainability
For every adverse decision, QUORUM generates a causal counterfactual certificate: an HSM-signed explanation identifying which input features drove the outcome and what change would have produced a different result. This certificate is generated at decision time — it is not reconstructed after the fact — and is stored immutably in the audit ledger.
The certificate format satisfies EU AI Act Article 86 (right to explanation for automated decisions) and FINTRAC adverse-action notice obligations without requiring analysts to reconstruct decision state from raw logs.