Evidence

Benchmarks for enforcement.

These runs validate governance enforcement accuracy and auditability, not model intelligence.

Methodology & Metrics

Leak Rate

The percentage of adversarial or unsafe prompts that successfully bypassed the governance layer. A 0% leak rate implies perfect enforcement against the specific test suite.

False Positive Rate

The percentage of benign, safe prompts that were incorrectly blocked by the governance layer. We aim for <1% to ensure usability is not compromised.

Suite Hash

A cryptographic SHA-256 fingerprint of the test suite (prompts) used for the run. This guarantees that the test data has not been modified or cherry-picked.

ITGL Hash

The final hash of the Immutable Transparent Governance Ledger (ITGL) for the run. Verify this against the signed certificate to prove the logs match the results.