Semantic Crosswalk and Type-Separation Canon Locks: Load-Bearing Terms, Phase Boundaries, and Non-Substitution Rules for the Applicability Boundary Doctrine

This record contains the semantic crosswalk and type-separation canon of the authored Applicability Boundary Doctrine line, together with its subordinate commercial-readability companion.Semantic Crosswalk and Type-Separation Canon Locks: Load-Bearing Terms, Phase Boundaries, and Non-Substitution Rules for the Applicability Boundary DoctrineVersion: 1.0Publication date: 2026-08-18Author: Vadym Partasyuk, Independent ResearchThis record consolidates the semantic architecture of the Applicability Boundary Doctrine corpus. It fixes the load-bearing lexicon, separates authority and phase types, records the non-substitution canon, and defines conflict rules for future publications in the line. It introduces no new doctrine, specifies no runtime enforcement, and introduces no controller.For avoidance of terminological confusion, "Applicability Boundary Doctrine" in this record denotes the authored AI-governance and consequence-boundary corpus. It does not denote the property-law agreed-boundary doctrine or the machine-learning concept of an applicability domain.The primary risk addressed is not the absence of vocabulary but silent substitution between terms that answer different questions. A term that has become load-bearing in one record can be quietly reused in another to carry a burden it was never defined to carry, and the substitution is difficult to detect because the vocabulary looks familiar. This record makes the substitution visible and names it as a defect.The central position is that legitimacy is a conjunction of distinct burdens rather than a scalar score. Correctness is not applicability. Similarity is not relevance, relevance is not source authority, and source authority is not contextual admissibility. Approval is not execution, and acknowledgement is not authority standing. A record, receipt, or replay is not standing. A documented control is not capacity until its end-to-end effect is measured against the remaining window. A score does not open a failed gate. An indicator, threshold crossing, or verifier finding is not a declaration, and a declaration is not a command.Public interpretation of this record depends on its own text and on explicitly deposited records. It does not depend on internal canonical amendments, unpublished working files, chat history, or generated synthesis. An internal amendment may not retroactively supply meaning to a term in this record; a change of meaning must be published as an explicit clarification or superseding record that identifies the exact semantic conflict.The record fixes the phase vocabulary from proposal through execution request, execution start, commit, consequence attachment, effect completion, and closure, together with the temporal extension rule under which a delayed consequence requires the relied-upon facts either to remain valid through attachment or to be re-evaluated before it. It fixes a load-bearing lexicon of twenty-four terms, each stated with what it does in this corpus and what it must not be confused with, and it retains Structural Authority Injection as a named class for unearned burden-bearing authority acquired through structural or control-plane position rather than through admissible grounds. It states the type-separation rules under which observation, applicability assessment, declaration, transition authorization, execution permission, enforcement, and record remain separately attributable, and under which a causal arrow between two types creates neither identity nor transfer of authority. It records a twenty-one-line non-substitution canon, seven cross-publication conflict rules, a lifecycle-assurance boundary, six interpretation states, integration controls, and a falsification surface of eight conditions under which the record is to be narrowed, revised, or withdrawn.The record declares its own contribution boundary explicitly. The formal burden structure of commit-time conjunction, the detailed consequence-bearing transition phase model with its closure obligations, the ordered attribution chain that assigns each type an ordinal position, the insurer-facing readability of deterministic liability, and the taxonomy of Boundary Event classes are each reserved to separate records. Two of the eight falsifiers test the new commitments this version introduces: that no public file of this record requires a non-deposited internal amendment in order to be interpretable, and that this record has not absorbed the contributions it reserves.Within this authored corpus the Applicability Boundary Doctrine is the single parent doctrine, and context admissibility, retrieval admissibility, transition semantics, execution-boundary profiles, and operative-capacity profiles are layers or instantiations of it rather than independent parent doctrines. This is a statement about the internal organisation of one authored corpus. It is not a claim of ownership over ordinary terms: applicability, standing, authority, commit, boundary, evidence, and proof are common words in engineering, law, and philosophy, and no exclusive right to any of them is asserted.Transaction-level applicability artifacts governed by this vocabulary are evidence inputs to lifecycle assurance. They do not replace hazard analysis, safety cases, conformity assessment, certification, or domain-specific verification and validation, and presenting a transaction-level artifact as though it discharged a lifecycle-assurance obligation is itself a substitution of the kind this record exists to name.The external-foundations section identifies established work and contemporary public adjacency before stating the narrower contribution retained. Safety and systems engineering, execution-time admission work, runtime authority reconstruction, pre-action authorization protocols, architectural governance frameworks, identity and authorization standards, legal and evidentiary disciplines, and indirect prompt injection research are each credited with what they already establish, together with the explicit non-claim that follows. Chronology and provenance are treated as record questions that resemblance does not settle: whether an identical phrase appears elsewhere, whether a concept is adjacent, which record is earlier, whether one author was exposed to another's work, whether one work architecturally depends on another, and whether anything is novel in the legal sense are six distinct questions, and this record answers none of the last three. No novelty is claimed merely because an exact phrase was not found in a search.An exact-title and distinctive-phrase similarity review was conducted on 2026-08-17 against the arXiv, Zenodo, DataCite, and Crossref public APIs and open-web search indexes. Subscription bibliographic databases, full-text search of the ACM Digital Library, patent databases, and non-public corpora were not searched, and no claim in this record depends on their coverage. The review is not a plagiarism score, a legal novelty opinion, a patent search, a freedom-to-operate analysis, or proof of independent creation.The twenty-three deposited records cited were verified individually against live repository metadata on 2026-08-17, and titles are reproduced as deposited rather than in short form. Where an earlier internal working list used a shortened or mislabelled form of a deposited title, the live deposited title governs and has been used. Bulk acceptance of a generated bibliography was not used.The second file in this record is Commercial Companion Note 10, Semantic and Type-Separation Integrity Readability. It is a subordinate commercial-readability companion for procurement, audit, board, counsel-facing, insurer-facing, safety governance, product-claim, interface and incident-review readers. It has no independent DOI and no independent doctrinal standing. It translates the governing record's type separation into nine claim types with the question each does and does not answer, twelve review questions, a seven-rank evidence order, a claim-to-type matrix, an authority and record attribution matrix, fourteen red flags, eleven corrective actions, a minimum review record, ten buyer-side questions, and eight review dispositions. It confers no certification method, legal opinion, safety case, or implementation design, and where it conflicts with a deposited defining record, the deposited record governs.A separately scoped Semantic and Type-Separation Integrity Review is available for organisations that require an auditable governing-source map, term and version ledger, claim-to-type matrix, non-substitution breach register, attribution matrix, reliance disposition and remediation plan for a defined document corpus or reliance decision. An optional Semantic Control Pack may be licensed for internal organisational use. These commercial services and private instruments are not part of the Zenodo deposit and confer no certification, legal opinion, safety approval or implementation right.This release should be read alongside the prior public publications in the Applicability Boundary Doctrine line, including the governing baseline DOI 10.5281/zenodo.19425317, Formal Conditions of Standing and Re-Sanctioning DOI 10.5281/zenodo.19599888, Unified Claim Standing Record Layer DOI 10.5281/zenodo.20479042, Authority Standing and Commit Standing DOI 10.5281/zenodo.20561344, Post-Consequence Proof Standing and Durable Reconstructability DOI 10.5281/zenodo.20573235, Operative Capacity Adjunct and Mission Temporal Validity DOI 10.5281/zenodo.20583953, and the immediately preceding Corpus Continuity and Canon Policy Record DOI 10.5281/zenodo.21983371, which governs channel precedence, citation practice, and publication ordering for this corpus.This release is domain-neutral and consequence-class based. It is relevant to consequence-bearing AI, high-consequence AI, agentic AI, clinical AI, AI/ML-enabled medical devices, banking and model risk, insurance, cybersecurity operations, critical infrastructure, avi