Runtime Execution Governance for AI Systems: A Cross-Platform Synthesis and Architectural Framework
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
AI governance is undergoing a structural transition from policy-based, pre-deployment oversight to execution-time control of autonomous and distributed systems. While prior work has addressed policy frameworks, guardrails, and monitoring, a coherent architecture that unifies governance intent, execution semantics, and system realization remains absent. This paper presents a cross-platform synthesis of research, engineering practice, enterprise systems, regulatory frameworks, and practitioner discourse, identifying a converging paradigm: governance as a runtime decision function over system state transitions. Building on this synthesis, we introduce a vertically integrated governance stack consisting of governance compilation, the Runtime Governance Architecture (RGA), the Artificial Intelligence Governance Control Plane (AGCP), and the PBSAI governance ecosystem. RGA formalizes governance as a deterministic execution model over state transitions; AGCP provides a control-plane implementation that enforces execution-bound authorization, lifecycle derivation, and invariant correctness; and PBSAI operationalizes governance as a domain-structured, evidence-centric multi-agent system. Governance compilation bridges policy intent to executable artifacts across these layers. We further provide a comparative systems analysis demonstrating that existing approaches---including policy engines, guardrails frameworks, and regulatory models---lack end-to-end alignment between intent, enforcement, and execution. The proposed stack resolves this gap by enabling deterministic, auditable, and non-bypassable governance at runtime. The results position runtime governance not as an extension of policy enforcement, but as a foundational architectural layer for AI-enabled systems.
Fields of Study & Subjects
Ethics and Social Impacts of AI
Access Control and Trust
Safety Systems Engineering in Autonomy
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 30 March 2026
- Language
- EN