Organizational AI Deployment Sovereignty: A Multidimensional Theoretical Framework for Governing Open- and Closed-Weight AI Systems

The rapid proliferation of generative artificial intelligence (AI) requires organizations to decide how AI systems are accessed, deployed, and governed. Existing research examines privacy, security, performance, compliance, and vendor dependence separately, without an integrated account of organizational control, dependency, and accountability. This study introduces Organizational AI Deployment Sovereignty (OADS), defined as the extent to which an organization maintains sustained, context-appropriate control over the data, model, infrastructure, runtime safeguards, and governance processes underlying AI deployment. Integrating Resource Dependence Theory, Sociotechnical Systems Theory, and Value Sensitive Design, and aligned with and informed by the NIST AI Risk Management Framework as an operational reference, this framework identifies five control dimensions, specifies observable indicators and ordinal assessment criteria for each, and treats them as a formative construct that is non-compensatory in high-risk contexts. It distinguishes OADS from digital sovereignty, organizational autonomy, vendor dependence, and AI governance. It particularly replaces the open–closed continuum with a multi-attribute configuration space and proposes two mechanisms: (a) the Sovereignty–Responsibility Paradox and (b) the Capability–Sovereignty Trade-off. Nine propositions are stated with mechanisms, boundary conditions, and disconfirming observations. A worked illustration and a supply chain application show how this framework identifies governance gaps and can change deployment decisions.