Static Permissibility in Dynamic Normative Environments: Dynamic Classification Failure as a Structural Property of Institutional Computation

This paper argues that computational systems deployed in consequence-bearing institutional contexts inherit an often implicit assumption: that the normative conditions under which an action was classified as permissible remain stable until the action is executed. The paper formalizes the failure of that assumption as Dynamic Classification Failure (DCF): reliance on a stale permissibility label in a normative environment that has changed between classification and consequence. The analysis defines DCF as a state-transition failure in institutional computation, distinguishes it from authorization defects, implementation errors, and design failures, and situates it within the Generalized Intentionality Mismatch Theorem. It derives necessary conditions for detecting and preventing DCF and connects the formal model to legal and regulatory frameworks for automated institutional action, runtime admissibility, AI governance, and agentic systems.