Every major cloud provider and emerging AI governance platform has converged on a similar architectural response to the agentic governance problem: a registry. Registries record what agents exist, what permissions have been granted, and what policies apply. They are necessary. They are not sufficient. This paper identifies the structural reason registries stall — the substrate beneath them is undefined — and describes the constitutive governance layer that makes registry-based enforcement coherent rather than merely descriptive. The distinction is not taxonomic. It is architectural, and it determines whether governance is a property of the system or a control imposed on top of it.
Narnaiezsshaa Truong (Wed,) studied this question.