Abstract

Public-sector adoption of artificial intelligence (AI) is accelerating, yet governmental use of AI unfolds within decision environments structured by legality, procedural fairness, privacy, equality, ministerial accountability, and democratic scrutiny. This review advances the concept of constitutional intelligence as a governance architecture for trustworthy public-sector AI. Using a narrative-comparative review of authoritative policy instruments, legal frameworks, and peer-reviewed scholarship from Canada, the United Kingdom, Europe, and major international standard-setting bodies, the article identifies a convergent set of design expectations for public AI: rights-sensitive scoping, ex ante impact assessment, documentation and traceability, human oversight, proactive transparency, and continuous institutional review. The analysis further shows that responsible AI adoption in government depends not only on model-level controls but also on the quality of data stewardship, interoperability, and administrative continuity. These findings are synthesized into a constitutional-intelligence model suited to Westminster-style institutions and other public administrations that must justify automated action under conditions of legal contestability and public accountability. The central claim is that constitutional and democratic requirements should not be treated as external compliance burdens. When embedded into institutional design, they operate as productive constraints that improve legitimacy, implementation discipline, and the long-term trustworthiness of AI-enabled public decision-making.