Governing AI with trust: an adaptive framework for institutional legitimacy in the UK public sector

Purpose This study aims to examine the limitations of artificial intelligence (AI) regulation in the UK public sector, particularly its fragmented and non-binding nature. It argues that current regulatory approaches lack the institutional coordination, legitimacy and transparency required to foster public trust in algorithmic decision-making. The paper proposes a conceptual model that reframes trustworthy AI not as a product of compliance or ethics alone but as the outcome of adaptive, legitimacy-centered governance. Design/methodology/approach The study uses a conceptual policy analysis approach, synthesizing literature from public administration, regulatory theory and AI governance. It critically assesses the UK’s “pro-innovation” regulatory model and develops a governance-oriented framework grounded in legitimacy, coordination and accountability. The framework is supported by illustrative cases from National Health Service AI applications and the GOV.UK algorithmic transparency initiative, with broader applicability discussed in relation to other public sector domains. Findings The analysis finds that non-binding, sector-led regulation in the UK lacks institutional alignment and accountability mechanisms, undermining public trust. The proposed framework reframes AI governance as a dynamic process of inter-agency coordination, transparent oversight and legitimacy production. Research limitations/implications As a conceptual paper, this study does not present empirical validation. However, it offers a testable framework for future research. The model can be adapted for comparative studies or case-based evaluation in other governance domains such as justice or finance, and it calls for the development of legitimacy indicators and enforcement mechanisms in AI policy. Practical implications This framework provides actionable guidance for policy designers, suggesting the need for institutionalized coordination, independent review bodies and legitimacy-based metrics for public sector AI oversight. It supports the design of governance models that go beyond technical compliance and embed trust and accountability into digital systems. Social implications By positioning legitimacy as a governance outcome, the framework underscores how AI policies should address not only risks but also public perception, equity and institutional behavior. It highlights the role of citizen engagement, redress mechanisms and transparency in sustaining democratic accountability in algorithmic systems. Originality/value This article makes an original contribution by framing AI governance as a public trust challenge and proposing a conceptual model rooted in legitimacy, institutional coordination and adaptive oversight. Unlike principle-based or compliance-driven approaches, the model bridges legal regulation and democratic accountability, offering a realistic, governance-centered alternative for the public sector.