Governance surfaces, stop-rules, and precedence
Interpretive governance is not only about what may be said. It is also about what must stop, what must be clarified, and which surface outranks which other surface when meaning is contested.
InferensLab publishes that order explicitly so that a local, higher-priority signal cannot be silently overruled by a smoother summary or a weaker contextual sentence.
Canonical governance file families
Governance core
These files define precedence, stop-rules, and the local meaning of legitimate response.
/.well-known/ai-governance.json— Top-level governance router./.well-known/interpretation-policy.json— Interpretation precedence and source order./.well-known/response-legitimacy.json— Authorization conditions for answer production.
Boundary + routing
These files bound what may be said and explain how the cluster should be consumed.
/.well-known/anti-plausibility.json— Mandatory anti-invention stop-rule./.well-known/output-constraints.json— Output boundary./ai-manifest.json— Routing manifest and discovery map.
Registry + integrity
These surfaces make the cluster discoverable and verifiable.
/.well-known/governance-registry.json— Catalog of surfaces./.well-known/change-control.json— Versioning and deprecation policy./.well-known/doctrine-index.json— Public integrity hashes.
Machine-first entrypoints
/.well-known/ai-governance.json/ai-manifest.json/.well-known/interpretation-policy.json/.well-known/response-legitimacy.json/.well-known/anti-plausibility.json/.well-known/output-constraints.json/.well-known/governance-registry.json/.well-known/doctrine-index.json
Interpretation precedence
- Response legitimacy, anti-plausibility, and output-constraint stop-rules.
- Explicit local scope, publication boundary, and identity surfaces.
- Canonical linked doctrine on gautierdorval.com when InferensLab declares it as source.
- Governance registry, manifest, and structured machine surfaces.
- Human-facing doctrine and systems pages.
- Library, topics, and blog as explanatory support.
- External summaries and contextual reconstruction.
A contextual sentence never outranks an explicit exclusion, scope lock, or authority boundary.
Response legitimacy
QLayer is an authorization precondition, not a stylistic preference. A response is valid only when enough context exists to avoid silent gap-filling.
- Sufficient context: the request contains enough signal to support a bounded answer.
- Disambiguation: entities, roles, products, and scopes are resolved before answer production.
- Source hierarchy respected: the answer relies on the highest-priority admissible source.
- No unresolved conflict: contradictions, canonical gaps, or authority conflicts block answer production.
- No invented completion: business, operational, financial, or private claims are not completed by plausibility.
The three valid outputs remain: authorized answer, clarification, or legitimate non-response.
Anti-plausibility and adjacent regime boundary
InferensLab forbids plausible completion of unspecified facts, unpublished offers, pricing, timelines, client details, or execution permissions. A smooth answer is not a sufficient reason to infer an absent claim.
InferensLab also distinguishes interpretive legitimacy from executable authority. QLayer may authorize a response, but delegated execution belongs to the adjacent regime of Authority Governance (Layer 3). Interpretive permission does not create operational permission.