Perspective proposes functional AI sovereignty framework based on conceptual synthesis, not empirical validation

A perspective published in Discover Artificial Intelligence proposes that governments can retain control over critical AI capabilities without owning the full technology stack. Its central finding is that effective sovereignty depends less on the nationality or location of assets than on public authorities’ capacity to govern dependencies through six interdependent functions: visibility, authority, allocation, substitutability, continuity and accountability. The article presents this as a theory-building framework rather than an empirically validated measurement model.

The author synthesises research on digital sovereignty, AI infrastructure, public-sector capacity, cloud concentration, procurement, interoperability, resilience and accountability. Contemporary policy initiatives are used illustratively rather than analysed as formal case studies. The synthesis is purposive: sources were selected for their relevance to conceptualising technological dependence, deriving the public functions needed for effective control and identifying institutional mechanisms for exercising those functions. No datasets were generated or analysed, and the article does not construct a country ranking or test a causal model of sovereignty.

The framework separates three forms of control that are often treated as interchangeable. Ownership is legally recognised possession of assets such as chips, servers, data centres, datasets and model weights. Operation is the technical and organisational capacity to deploy, maintain, update and recover systems. Governance is the enforceable capacity to determine the conditions under which critical capabilities are supplied and used. The perspective argues that ownership without expertise, software access or upstream resources may provide only nominal control, while privately or foreign-supplied infrastructure may remain compatible with sovereignty when its dependencies are transparent, disciplinable and reversible.

Within this approach, “functional AI sovereignty” denotes the durable, legitimate and accountable capacity of public authorities to govern the dependencies through which critical AI capabilities are produced, accessed, allocated, substituted and sustained. Visibility requires authorities to identify and interpret material dependencies; authority requires enforceable decision rights; allocation concerns the ability to direct scarce resources toward publicly determined purposes; substitutability requires credible and timely replacement options; continuity concerns preservation of public functions during disruption; and accountability requires consequential public and private decisions to be reviewable and contestable. These capacities are mutually dependent rather than a checklist of independently desirable attributes.

The article introduces “sovereign orchestration” as an institutional model for applying this framework across distributed resources. Its proposed mechanisms include regulation, strategic procurement, interoperability, multi-provider routing, capacity clearing and selective public ownership. Unlike nationalisation, orchestration would not transfer the entire stack into public hands; unlike conventional outsourcing, it would establish a common control plane over participation, information access, allocation, substitution and continuity. Its practical test is whether authorities can recompute the delivery of a critical capability when a model, cloud service, data environment or other component becomes unavailable or unacceptable.

Proposed governance mechanisms include disclosure and audit rights, continuity requirements, portability provisions, migration assistance, public-priority access and independent review. Routing could direct workloads among accredited providers according to capability, security, jurisdiction, cost and other authorised criteria. A capacity-clearing mechanism could meter and settle public allocations across providers where entitlements must be portable or reserved capacity redistributed. The perspective recommends institutional separation among policy-setting, provider accreditation, technical operation and complaint adjudication so that one actor does not simultaneously control the rules, compete as a supplier and adjudicate disputes.

The model has explicit limits. Orchestration is viable only above what the article calls an “orchestration viability frontier,” requiring sufficient public administrative and technical competence, informational access, credible market and infrastructural alternatives, and the capacity to revise arrangements as technology and supply chains change. These requirements are capacity-contingent and cannot be assumed to transfer unchanged from European and North American jurisdictions to states with more limited infrastructure, compute, skills or regulatory resources. Shared regional expertise, pooled procurement and foundational public investment may be more realistic starting points in lower-capacity settings.

Direct public ownership remains necessary under an “ownership trigger” when three conditions converge: interruption would impose unacceptable harm, no technically and legally independent substitute can be activated within the relevant period, and continued access is vulnerable to an external legal or geopolitical authority beyond effective domestic control. The appropriate object is a minimum sovereign capability needed to preserve the essential public function, not necessarily frontier-scale compute or a complete national stack. Even then, public ownership should be proportional rather than treated as a presumption of permanent state provision.

The perspective also identifies a “sovereignty paradox”: efforts to reduce external dependence can produce infrastructural recentralisation, protectionist lock-in, surveillance expansion or technocratic displacement. It therefore proposes “dual contestability,” under which private suppliers must remain challengeable when they exercise infrastructural power and public authorities must remain accountable when allocating capacity, restricting access or invoking emergency powers. The framework’s contribution is an analytical architecture for comparing forms of AI dependence; its limitations include the absence of empirical validation, the dependence of institutional design on public capacity, and the need for comparative research across jurisdictions, sectors and layers of the AI stack.

Subject of Research: Functional AI sovereignty and sovereign orchestration for governing critical AI dependencies

Article Title: Governing critical AI capabilities through sovereign orchestration without full stack ownership

Article References: Hao, Y. (2026). Governing critical AI capabilities through sovereign orchestration without full stack ownership. Discover Artificial Intelligence, 6(1), Article 1440. https://doi.org/10.1007/s44163-026-02447-x

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02447-x

Keywords: AI sovereignty, critical AI infrastructure, sovereign orchestration, functional sovereignty, public procurement, cloud concentration, interoperability, provider substitution, institutional capacity, accountability, Governing, critical