Sovereign Software

Possessing application source does not by itself establish control over the system that gives the application its behavior. Dependencies, build tools, runtimes, identity services, update channels, remote infrastructure, and hardware can retain powers that the nominal owner neither understands nor governs. This paper develops sovereign software as a relational and operational property: the ability of identified people or institutions to determine software meaning, constrain effects, inspect the grounds of assurance, and continue, repair, or replace required capabilities without ungranted control by an indispensable supplier. As a companion to Constitutional Computing: Evidence-Carrying Realization, Agentic Development, and the Economics of Structured Abundance, it presents sovereignty as the purpose that structured abundance can help make practical. It distinguishes legal freedom, source access, custody, security, and effective independence; develops a thirteen-axis custody profile and a claim-relative dependency-closure model; and treats reconstruction and provider-removal exercises as tests of meaningful exit. Constitutional contracts connect these requirements to independent evidence, locally governed adoption, and competing implementations. In universal-GSM terms, the paper follows sovereignty from possibility and specification through admissible and preferred configurations, a selected instance constitution/configuration, its instance/implementation realization, evidence, and revision. Agentic development and distributed publication are proposed as ways to turn opaque dependencies into bounded reliance relationships and, where feasible, replaceable realizations. Neither an attestation nor a complete inventory confers sovereignty by itself. The paper examines build bootstrap, information-flow limits, version sovereignty, closed-source adaptation, shared governance, and the risks of creating new gatekeepers around models or verification. An encrypted-messaging scenario and a falsifiable evaluation program specify how the approach could be investigated without claiming universal verification, zero residual trust, or a mature sovereign operating system.