Yu, Tak Ming, Zheng, Xia and Ng, Hoi-kuen (2026) From responsibility gaps to meaningful human control: an ethical and regulatory governance framework for intelligent AI systems. AI and Ethics. ISSN 2730-5953 (In Press)

PDF/A (Will be made OA based on IRR on publication) YuT_From responsibility gaps to meaningful human control_AM_PDFA.pdf - Accepted Version Restricted to Repository staff only Download (320kB)

Abstract

Artificial intelligence (AI) systems are increasingly capable of influencing consequential decisions, creating concerns about how human responsibility can be preserved when system behaviour becomes more autonomous and difficult to understand. Although responsibility does not disappear, it may become distributed and obscured across designers, developers, deploying organisations, operators, supervisors and public authorities. This creates a governance challenge: ensuring that responsibility remains identifiable, practically exercisable and institutionally enforceable. This study examines how Meaningful Human Control (MHC) can address this challenge through a system-sensitive governance approach. It introduces two complementary analytical frameworks. The Four AI System Model classifies AI systems according to autonomy and intervention, identifying different governance configurations and the forms and intensities of control they require. The Four-Layer Governance Model provides a common governance architecture across conceptual, operational, legal and institutional arrangements, while recognising that the priorities and mechanisms of control within each layer vary according to autonomy, intervention and associated risks. Using police internal monitoring and facial recognition as contrasting examples, the study demonstrates how MHC can be translated from ethical principles into governance practice. The analysis shows that meaningful control depends not only on human involvement in decisions, but also on defined purposes and boundaries, operational procedures for review and correction, legal safeguards, and capacity for oversight and intervention. The study argues that responsible AI governance requires more than technical explainability or formal human oversight. Governable AI requires bounded authority, traceable responsibility, meaningful human control, and capacity to challenge, correct or suspend decisions when necessary.

Item Type:	Article
Additional Information:	This version of the article has been accepted for publication, after peer review (when applicable) and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at:
Subjects:	Computing > Intelligent systems Social sciences
Date Deposited:	10 Sep 2026
Dates:	Date Publication status 31 August 2026 Accepted
School, department or research centre:	School of Human and Social Sciences
URI:	https://repository.uwl.ac.uk/id/eprint/15304
Sustainable Development Goals:	Goal 9: Industry, Innovation, and Infrastructure	Sustainable Development Goals:	Goal 10: Reduced Inequalities	Sustainable Development Goals:	Goal 16: Peace, Justice, and Strong Institutions

Actions (admin access)