OTANIS: An Operational Trust and Authority Normative Integrated System for Executable Governance of Agentic AI Across Multiple Irreversibility Boundaries

As AI systems evolve from bounded inference engines into live agentic architecturescapable of initiating irreversible actions, governance failures increasingly occur despite localcompliance of individual components. These failures are structural rather than behaviouraland arise when legitimate authority cannot be demonstrated or enforced at execution time.This paper introduces OTANIS, an Operational Trust and Authority Normative IntegratedSystem, as a unified architectural governance framework for critical agentic systems thatproduce irreversible outcomes. OTANIS integrates ex-ante admissibility, runtime authorityenforcement, authority lifecycle semantics, compositional preservation, deterministic bindingto the earliest governed irreversible execution boundary, conflict escalation, and multi-layercross-domain governance into a single system-level architecture. It is designed to governaction paths with multiple candidate execution boundaries and to preserve authority acrosscomposed systems spanning multiple domains and governance layers. Authority is treated asa first-class executable object with explicit validity, revocation, refusal, fallback, provenance,traceability, accountability, and audit requirements enforced at the bound execution boundary.The framework is model-agnostic, falsifiable, and scoped specifically to live agentic systemsrather than offline or purely advisory AI. Formal definitions, deterministic boundary binding,atomicity semantics, snapshot and replay requirements, provenance requirements, boundedlatency handling, and integrity-based audit criteria are provided. The paper also explains whatthis architecture means in practice: it creates a basis for designing governable agentic systems,reviewing structural authority enforcement before deployment, assessing procurements forruntime governability, and auditing irreversible action paths after execution. Its guaranteesare enforcement-correct relative to declared governance artefacts and declared irreversibilitysurfaces; it does not infer omitted boundaries, omitted dependencies, or incomplete authoritypredicates. OTANIS is proposed as a unified reference architecture for the design, review,procurement, and audit of critical agentic AI systems producing irreversible outcomes.