Governance Geometry: What Execution Integrity Cannot Govern

Current discourse on AI execution integrity treats the hardware boundary as the primary locus of governance. This paper argues that execution integrity, while necessary, is insufficient as a standalone architecture. Sovereignty in multi-stakeholder AI systems requires governance structures capable of distinguishing upstream distortions from downstream enforcement. This work identifies the structural surfaces missing from boundary-centric approaches and outlines the conditions any governance model must satisfy to remain valid across heterogeneous actors, models, and temporal contexts.