Whose AI Runs the Government? Infrastructure, Sovereignty, and the Case for Transparent Public AI

This paper argues that the decisive AI sovereignty question is: who owns, governs, and controls the AI systems increasingly integrated into the everyday machinery of the state: tax administration, health records, social benefits, immigration processing, regulatory enforcement, and municipal service delivery. Most Western democracies are deploying AI across these functions on foreign-owned cloud infrastructure, under foreign legal jurisdiction, with limited transparency and no coherent strategy for how deployment at the federal level connects to deployment at the provincial, municipal, or treaty-alliance level. The paper identifies this gap as the “sovereignty leak,” the structural vulnerability created when national sovereignty commitments dissolve into foreign dependency at sub-national levels of government. It introduces the Sovereign AI Maturity Model, a three-axis assessment (Infrastructure Sovereignty, Policy Maturity, Application Depth) that reveals the most dangerous configuration to be deep AI integration combined with weak sovereignty and weak governance. Eight models of sovereign AI strategy now emerging globally are analysed, from the United States and France (frontier builders) to China (AI as state infrastructure), Japan (national consortium), Singapore (adapted open-weight), Switzerland (public infrastructure AI), India (digital public infrastructure sovereignty), and South Korea (competitive national championship), alongside Canada’s position as a case study in dependent sovereignty. The paper proposes five minimum transparency standards for government AI, a procurement reform framework that introduces sovereignty as a weighted criterion, and an executive roadmap for Tier 2 nations with a two-year action window. An accompanying Sovereign AI Scorecard provides a nine-indicator assessment tool for evaluating the sovereignty posture of any government AI deployment. First in a three-part series on AI sovereignty. The companion papers address the military dimension and the emerging cost structure of sovereign AI infrastructure.