Responsible AI Governance in Banking: Integrating Model Risk, Consumer Protection, and Operational Resilience Across the United States, European Union, United Kingdom, Singapore, and Hong Kong

Banks are embedding artificial intelligence (AI) into credit underwriting, fraud monitoring, trading, collections, customer service, and anti-financial-crime operations. The resulting risk profile is no longer captured adequately by a single governance lens. A bank AI system can simultaneously be a model-risk object, a consumer-impact mechanism, and a technology dependency whose failure may disrupt critical services. This paper develops an integrated Responsible AI governance framework for banks through a comparative documentary analysis of primary regulatory and supervisory materials across five influential jurisdictions: the United States, the European Union, the United Kingdom, Singapore, and Hong Kong. The analysis shows a convergence of supervisory expectations around three control families: lifecycle model governance, customer protection and explainability, and operational resilience with third-party accountability. Building on this convergence, the paper proposes an integrated control architecture covering governance bodies, AI inventory and use-case classification, risk tiering, validation, explainability, fairness review, deployment controls, incident response, vendor oversight, and ongoing monitoring. Illustrative case studies show that fragmented governance often misses failure modes even where no single regulation has yet been formally breached. The paper concludes with a phased implementation roadmap and argues that banks should govern AI as a cross-cutting enterprise capability rather than as a narrow model-development topic.

Responsible AI; banking; model risk management; consumer protection; operational resilience; AI governance; AI Act; DORA; SR 11-7; FEAT; HKMA