Does Cybersecurity Influence the Impact of AI on Bank Risk-Taking? Evidence from Dual-Banking Countries

Abstract

Using 5,806 bank–year observations from 17 Asian and African economies over the years 2012–2022, we examine how artificial intelligence (AI) adoption influences bank risk-taking and whether cybersecurity capacity moderates this relationship. We find that AI intensity is associated with higher risk-taking at prevailing adoption levels.  We also note that their relationship is concave, suggesting a shift from “risk-ramping” during early deployment to “discipline” as model governance and monitoring mature. We also find that stronger cybersecurity attenuates AI’s marginal risk effect. Heterogeneity is evident: conventional banks exhibit higher turning points, reflecting a longer risk ramp, whereas Islamic banks peak earlier, consistent with stricter governance structures and more risk-averse practices. Results are robust in various sensitivity analyses. The findings suggest that AI scaling in banking requires synchronized advancement in cybersecurity and a model-risk management framework, aligned with evolving supervisory doctrine on digital resilience and AI governance.