Artificial intelligence is increasingly embedded in core banking processes, raising significant challenges for prudential and conduct supervision, particularly in relation to model risk, governance responsibilities, documentation standards, and the enforceability of accountability obligations. This article provides a comparative analysis of how nine major jurisdictions—the European Union, United Kingdom, United States, Canada, Japan, Australia, India, China, and the Islamic banking states —supervise and regulate the use of AI in banking through risk-based and accountability-oriented oversight frameworks. It develops a ten-dimension comparative matrix to assess the degree of normative consolidation and institutional maturity across micro (individual rights), meso (institutional accountability), and macro (systemic safeguards) levels of supervision. The analysis identifies three archetypes of algorithmic governance—legal-normative, executive-fragmented, and state-centralized—and demonstrates that many jurisdictions adopt the language of “trustworthy” AI without establishing equivalent mechanisms of supervisory enforceability. The article argues that regulatory legitimacy in AI-driven banking depends less on the proliferation of high-level principles than on their translation into enforceable supervisory practices through proportional oversight, defined as the calibration of regulatory scrutiny to the impact of algorithmic systems on rights, prudential soundness, and systemic stability. The findings provide a structured basis for strengthening prudential, conduct, and operational-risk supervision in AI-intensive banking environments.
García-Llorente et al. (Thu,) studied this question.