Artificial Intelligence Governance at the Global Level: Policy, Standards and Institutional Design

Abstract

Article examines global governance of AI as an emerging polycentric system formed largely through soft-law instruments, technical standards, and institutional coordination mechanisms. It advances the concept of soft polycentricity to describe a governance condition in which multiple centers of rule-making possess practical authority but lack a binding hierarchy or stable coordination procedures. The study compares major regulatory approaches and assesses the International Telecommunication Union (ITU) as a nodal institution whose comparative advantage lies in technical standardsetting, multistakeholder convening, and capacity-building. The article also evaluates the structural limits of ITU authority, including mandate boundaries, geopolitical contestation, resource constraints, and possible institutional competition within the United Nations system. It argues that ITU should not be treated as a universal AI regulator, but as one pillar of a distributed UN-centered architecture capable of reducing fragmentation through standards coordination, capacity development, and procedural integration. The conclusion proposes measurable implementation steps for enhancing UN coordination in AI governance while preserving institutional specialization.