Banking resilience in the era of dual AI risk-adoption gap analysis and strategic implications for global AFC

Abstract

This paper examines how the global banking sector preserves resilience under a dual technological shock: the criminal use of generative artificial intelligence (GenAI) and agentic AI1 in money laundering, fraud, and sanctions evasion, and the slow deployment of defensive AI/machine learning (ML) in anti-money laundering (AML)/anti-financial crime (AFC) programmes. It tests the hypothesis of a systemic ‘risk-adoption gap’ between the speed of threat evolution and the industrialisation of defensive AI. A qualitative triangulation integrates (1) the Association of Certified Anti-Money Laundering Specialists (ACAMS) Global AFC Threats Report 2025;2 (2) quantitative evidence from ‘Global AML Research: The Road to Integration’ (SAS‒ACAMS‒KPMG, 2024); (3) Financial Action Task Force standards on digital transformation and virtual assets; and (4) Europol and Cybercrime Trends 2025 analyses on AI-enabled cybercrime. Cross-source patterns inform policy implications. The study finds that only 18 per cent of institutions have fully operational AI/ML in AML/counter-terrorism financing, while 40 per cent lack any adoption plan, despite AI-driven threats rated high/very high. Legacy IT and data constraints and a talent gap in data/ML/explainable AI (XAI) are the main internal barriers. Asia combines high AI use with fragile infrastructures, whereas the US/Oceania show delays linked to regulation and core-system complexity. Offensively, GenAI amplifies authorised push payment and document fraud, synthetic identities and multi-channel phishing, eroding rule-based controls. Non-homogeneous data sources limit statistical generalisation but enable framing AFC 5.0 as asymmetric warfare, where AI powers both crime and defence. The paper formalises the risk-adoption gap as a driver of systemic stability, aligns AI-enabled threats, banks’ AI/ML maturity and regulation, and outlines a resilience architecture grounded in data, governance, and talent. This article is also included in The Business & Management Collection which can be accessed at https://hstalks.com/business/.