Abstract—The integration of artificial intelligence into banking has increased reliance on decision systems whose internal logic may not be fully reconstructible. Existing analysis has addressed opacity primarily at the institutional level, focusing on governance, explainability, and model oversight. This paper introduces Systemic Opacity Risk (SOR) as a macroprudential dimension of AI-driven finance. SOR is defined as the risk that the aggregation and correlation of institutional opacity — even if individually within tolerance thresholds — impair the financial system’s collective reconstructibility under conditions of systemic stress. The framework distinguishes SOR from model, operational, and cyber risk, and identifies structural mechanisms through which opacity may acquire systemic relevance, including model homogeneity, infrastructure concentration, and governance convergence. The analysis does not propose quantification or regulatory intervention. It articulates the conceptual conditions under which reconstructibility may function as a component of systemic resilience. The paper advances SOR as a structural hypothesis intended to inform macroprudential analysis in increasingly algorithmic financial environments. Third conceptual addendum to the Five Beacons Model as developed in the monograph: Garcia-Maceiras, JM (2026). 'The Banking Risk of AI Explanation: The Five Beacons Model'. Zyphrum ADR Notebooks No. 1, after Garcia-Maceiras, JM (2026) "The Five Beacons Model: A Prudential Architecture for AI Explainability and Legal Liability in Banking". Zyphrum ADR Notebooks No. 2 and Garcia-Maceiras, JM (2026) "Tolerance for Opacity: A Threshold Framework for AI-Driven Banking ". Zyphrum ADR Notebooks No. 3.
JM García-Maceiras (Thu,) studied this question.