Beyond Admissibility: Governance Maintenance in Long-Lived Autonomous Systems

Contemporary approaches to the governance of autonomous systems treat admissibility — the demonstration that a system may legitimately operate under a given authority structure — as a terminal event: design, validate, deploy, trust. This paper argues that this assumption is structurally false. An admissibility demonstration is performed over a concrete authority structure; because that structure can change during the operational life of the system, the validity of the demonstration can be lost during the operational life of the system. Admissibility is not a property a system possesses, but a condition that must be actively preserved. The paper develops this argument in four steps. First, it identifies the precise object of an admissibility demonstration: not the system's components or behavior, but the relational structure of authority over which the demonstration was performed. Second, it derives a non-circular notion of structural invariants as the operational representation of the conditions the demonstration requires to remain true. Third, it states an impermanence result — no demonstration of admissibility can demonstrate future admissibility — together with its corollary: a change is constitutionally material if and only if it alters some relation in the validated authority structure. Fourth, it characterizes a minimal relational basis (authorization, separation, precedence, scope) derived as a decomposition of authority specification itself, and identifies capability–authority conflation as the canonical failure mode by which admissibility erodes without any individually alarming change. A minimal formal model is given in which conditions of validity, constitutional materiality, constitutional equivalence, and the audit problem are each definable over a single finite relational structure, and a worked example shows the materiality test executing mechanically over concrete changes. The institutional consequence is a distinct operational role — the constitutional auditor — whose object of verification is relations between roles, not implementations, components, or outputs. Governance, on this account, does not produce governed systems; it produces systems that can remain governable. Part of the AMO (Authority-Mediated Operations) Research Series. This paper extends the admissibility results of the series to the full operational lifecycle, opening the operational-governance cluster of the research program. ---