Boundary-to-Bridge Constructive Doctrine for Human–AI Governance: Authority, Graph Evidence, Provenance, and Trust-Root Prelock

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This working paper and technical release package presents the terminal prelock state of BIT-AI-HUMAN-CONSTRUCTIVE-DOCTRINE-001, a Boundary Information Theory (BIT) framework for turning separation firewalls into typed constructive bridges between human reference semantics, AI authority, evidence, action, and institutional control.\n\nThe architecture contains three families of bridge contracts—Human-to-Normative (HN), Lineage/Capability-to-Authority (LA), and Knowledge/Power-to-Action (KP)—together with a common Authority/Grant Graph, Dependency/Composition/Beneficial-Control Graph, and Temporal Epoch/Expiry/Revocation/Replay kernel (KAG/KCG/KTE). It further introduces a Graph Evidence Certificate (GEC), an Evidence-of-Evidence Receipt (E2ER), and a Trust-Root Admission, Rotation & Recovery kernel (TRAR).\n\nThe terminal audit freezes the claim ceiling and records a 3,041/3,041 finite regression spine. These are finite architecture and regression receipts, not independent empirical observations and not proof of real-world AI safety. The work does not claim to solve alignment, instrumental convergence, containment, strategic deception, universal human values, legal or political legitimacy, world-graph completeness, or real-world trust-root security. No proof-assistant verification, runtime deployment, or whole-system empirical validation is claimed. The release is intended as a reproducible handoff for critique, formalization, adversarial testing, and future independent validation.