PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
March 26, 20260 citationsOpen Access

ARETABA: A Formal Execution-Time Authority Framework for Constructing and Enforcing Governed Irreversible Actions in Agentic AI Systems

View Full Paper
MOMasayuki Otani

Key Points

  • This research aims to formalize a framework for systematically governing irreversible actions in agentic AI during execution.
  • Isolates and formalizes ARETABA as an execution-time authority framework.
  • Introduces runtime control surface and construction discipline for authority objects.
  • Defines authority scope, boundary definition, and lifecycle validity based on governed ex-ante artifacts.
  • Presents a synthesis function to derive authority templates from ISDAIRE-governed action classes.
  • Demonstrates completeness, consistency, and executability criteria for ARETABA templates.
  • Shows that ARETABA supports boundary-evaluated admissibility used in OTANIS.
  • Presents a detailed example illustrating the application of ARETABA in a runtime-enforceable authority structure.

Abstract

As agentic AI systems move from advisory assistance to execution-bearing architectures,governance must remain enforceable at the moment a system-controlled action becomesirreversible. OTANIS defines a boundary-evaluated permission architecture and ISDAIREdefines ex-ante admissibility, but neither by itself answers the narrower construction questionof how runtime authority is systematically built, bound, and resolved for a governed actioninstance. This paper isolates and formalises ARETABA as the execution-time authorityframework that answers that question. ARETABA is presented as both a minimal runtime control surface and a constructiondiscipline for authority objects at the earliest governed irreversible execution boundary𝑇*𝑒(𝑎). The framework specifies how authority scope, boundary definition, lifecyclevalidity, revocation, refusal, escalation, traceability, accountability, and boundary-evaluatedadmissibility are derived from governed ex-ante artefacts, versioned boundary registries, anddeclared dependency scope. A class-level synthesis function is introduced to derive authoritytemplates from ISDAIRE-governed action classes, together with a runtime resolutionfunction that instantiates boundary-resolved authority objects for concrete action instances.The paper formalises completeness, consistency, and executability criteria for ARETABAtemplates and shows how these criteria support the boundary-evaluated admissibilitypredicate already used in OTANIS. The framework is intentionally fail-closed, replay-oriented,and bounded to action classes whose candidate boundaries, ordering semantics, dependencyscope, and controlled mediated commit paths can be declared ex ante and checked at runtime.It does not infer omitted hazards, omitted dependencies, or semantically incomplete authoritycontent after deployment. A detailed insurance payout example shows how ARETABA turns an admissible actionclass into a runtime-enforceable authority structure and where failure remains a specificationproblem rather than an execution-boundary enforcement problem. The paper positionsARETABA as the execution-time authority discipline underpinning OTANIS, complementingISDAIRE at the design layer and GAG/MGAG at the compositional layer.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Masayuki Otani (2026) studied this question.

synapsesocial.com/papers/69c4cd05fdc3bde448918ccehttps://doi.org/10.5281/zenodo.19199917
Ask AI
Helpful
Bookmark
Share
View Full Paper