Deterministic governance guarantees that identical governed state produces identical outcomes. That guarantee assumes governed state already exists. This paper examines the step that transforms unstructured reality into governed state and argues that the step itself must be governed. The paper introduces representation governance as the runtime mechanism through which governance ontology and constraint codification become governed state suitable for deterministic evaluation. It does not propose a new dimension of deterministic governance. Instead, it provides an operational reading of the existing Multi-Dimensional Deterministic Governance Model, making explicit a connective the model already implies. A central observation follows: Deterministic decision logic guarantees identical outcomes for identical governed state. It does not guarantee that identical real-world phenomena produce identical governed state. The paper argues that deterministic governance requires not only deterministic evaluation but also deterministic-or-fail-closed, replayable production of governed state. Where representation cannot be produced under those conditions, instability resolves to ABSTAIN rather than unsupported execution. The discussion is vendor-neutral and implementation-neutral. It focuses on governance ontology, constraint codification, governed state production, replayability, authorization, provenance, and evidentiary integrity. The paper positions representation governance as the operational bridge between ontology and codification on one side and deterministic decision logic on the other. Keywords: AI governance, deterministic governance, representation governance, authorization, execution-time authorization, governed state, ontology, constraint codification, provenance, replayability, proof-carrying decisions, AI safety, policy enforcement, runtime governance, auditability.
Meyman et al. (2026) studied this question.