CAIN-42 invariant · E32

E32-KREV-improvement: invalidated evidence propagates to improvement

Held · CAIN-42 Evolution 32 -- Governed Autonomy Learning Fabric

Last reviewed 2026-10-01

held   niche Identity, authority & delegation · family knowledge_revocation

What this rule means

CAIN-42 must always satisfy: invalidated evidence propagates to improvement. It is one of 303 invariants checked for CAIN-42 Evolution 32 -- Governed Autonomy Learning Fabric. An invariant is a rule the system may never break, whatever an agent or attacker does; the test suite tries to break it across many scenarios and records the result.

Recorded detail

{'observation': 0, 'action': 0, 'state': 0, 'prediction': 1, 'outcome': 0, 'error': 0, 'policy': 1, 'authority': 0, 'proof': 0, 'incident': 1, 'improvement': 1, 'experiment': 1}

Verify it in your browser

Your browser downloads the bundle's SHA256SUMS manifest and the file(s) behind this page, hashes them with SHA-256 locally (WebCrypto), and compares. A match shows the record you are reading is the published one; it does not by itself prove who published it (see the signed claims registry and the bundle verifier for that).

Check it yourself

Browse the raw bundle · How to reproduce it · SHA-256 manifest

Related rules

• E32-CFG-deny_prefixes: adding a deny_prefixes restriction is never authority drift
• E32-CFG-secret_patterns: adding a secret_patterns restriction is never authority drift
• E32-CFG-rate_limit: adding a rate_limit restriction is never authority drift
• E32-CFG-world_model_gate: adding a world_model_gate restriction is never authority drift
• E32-PROT-evaluator: evaluator is outside what learning may change
• E32-PROT-hidden_evaluation: hidden_evaluation is outside what learning may change
• E32-PROT-verifier: verifier is outside what learning may change
• E32-PROT-rollback: rollback is outside what learning may change

← E32-KREV-prediction · all 303 · E32-KREV-policy →