Toward a Multi-Level Architecture for AI Governance: Synthesizing Systematic AI Governance Literature and International Governance Pathways for Advancing AI

Abstract

Artificial intelligence governance is increasingly shaped by two partially disconnected conversations: one focused on responsible AI practices, standards, and governance frameworks across organizational settings, and another focused on the international governance of advanced or transformative AI under conditions of geopolitical competition. This paper synthesizes two recent publications representing these strands. The first is a systematic literature review of AI governance that analyzes 28 studies and organizes governance artifacts across team, organization, industry, national, and international levels. The second develops a taxonomy of international governance options for advancing artificial intelligence, emphasizing governance by different actors and intervention at the stages of development, proliferation, and deployment. Read together, these works suggest that effective AI governance must be conceived as a multi-level, stage-sensitive architecture. This paper argues that AI governance should combine responsible AI mechanisms at lower levels with stronger national and international controls for high-risk, dual-use, and transformative AI. The paper clarifies the complementarity of the two source publications, builds an integrated framework around the questions of who governs, what is governed, when governance intervenes, and how governance is implemented, and develops a research and policy agenda focused on governance alignment, chokepoints, competitive dynamics, and international coordination. The central conclusion is that no single governance mechanism can handle the full AI governance problem. Durable governance will require coordination among organizations, standards bodies, states, and international institutions, with governance intensity increasing as capabilities, risks, and strategic externalities grow.