This article argues that the governance of artificial intelligence (AI) requires a fundamental institutional redesign rather than incremental regulatory reform. Existing frameworks fail due to structural misalignment with AI’s autonomy, compositionality, and concentration dynamics. In a context where global coordination is infeasible and national regulation insufficient, the paper advances coalition-based digital governance as a second-best yet effective solution. It develops a comprehensive architecture grounded in Control?Accountability?Protection (CAP), operationalized through interoperable standards, data-control infrastructure, and coordinated enforcement. The article further proposes layered guardrails―including trade-based compliance, soft-power monitoring, and hard-power sanctions―to ensure enforceability and redress across jurisdictions.