Closing the Principles-to-Practice Gap: Operational AI Governance for Sustainable Digital Infrastructure

Abstract

Artificial intelligence is becoming part of public administration, critical services, transport, energy systems, and other forms of digital infrastructure. Yet many governance programmes still rely on broad commitments to fairness, transparency, privacy, and accountability without specifying how those commitments should be implemented, tested, and evidenced. This review examines the transition from principle-based governance to operational assurance. It draws on the thematic findings of a large systematic synthesis that retained 95 high-quality studies and frameworks published during the rapid expansion of AI governance research from 2020 to mid-2025. The literature shows a clear imbalance: privacy and ethics receive substantial attention, while operational security, bias control, accountability mechanisms, resilience, automation, and implementation costs are handled inconsistently. The review develops a practical interpretation of a three-layer governance architecture built around principles, controls, and evidence. It explains how organisations can translate values into technical and managerial safeguards, then demonstrate performance through logs, impact assessments, testing reports, monitoring records, and audit trails. The analysis also considers the relevance of this architecture to Sustainable Development Goal 9, particularly the need for reliable, inclusive, and resilient digital infrastructure. The central argument is that trustworthy AI cannot be achieved by publishing ethical principles alone. It depends on a repeatable operating model that assigns responsibility, matches controls to risk, collects verifiable evidence, and improves continuously across the AI lifecycle.