No-Proof No-Action Gate
No-Proof No-Action Gate means that a high-risk AI system must enter a visible no-action state when required proof, authority, or scope is missing.
Concept Index
Stable names for public evidence.
Each term has a fixed definition, evidence scope, DOI/GitHub/HF links, counterexample route, and boundary statement.
Self-Certification Grounding Check
Self-Certification Grounding Check means that a system's self-report must be grounded in public records, failure states, unresolved states, expiry conditions, and external constraints.
Proof-Carrying Action
Proof-Carrying Action means that an action claim must carry public proof fields, authority scope, refusal records, evidence boundaries, and counterexample routes before it can be trusted.
WisdomBench
WisdomBench is a public benchmark line for evaluating longitudinal failure learning, recovery, grader drift, and wisdom-oriented behavior under bounded evidence.
Robot/Vision Downgrade Rule
Robot/Vision Downgrade Rule means that uncertain perception evidence must route to review, stop, or wording that exactly matches the validated perception envelope; it must not become deployment language.
The Grounding Canon
The Grounding Canon is the P46 public line for reporting claim-to-evidence ladders across theory, constructed witnesses, shadow evidence, and human-reviewed records.