At the Credit 2026 workshop, Giuseppe Siani, the Bank of Italy’s Head of the Directorate General for Financial Supervision and Regulation, said the Bank has asked directly supervised intermediaries to assess their exposure to risks from frontier artificial intelligence models. Firms must review governance, cyber hygiene, ICT resources, vulnerability management, monitoring, defenses and resilience testing, then submit plans identifying responsibilities, investments and remediation measures by the end of 2026. Siani also outlined the gradual extension to less significant institutions of the supervisory calendar approach for covering nonperforming loans originated before 2019. A Bank of Italy review of AI systems used in creditworthiness assessments has preliminarily identified the need for stronger board oversight, AI-specific risk indicators and comprehensive rules covering model development, validation, use and monitoring. Firms also need stronger data governance, information systems and staff training. These priorities reflect wider operational risks identified through Digital Operational Resilience Act monitoring, including an 80% increase in IT incidents involving Italian intermediaries in 2023-25 compared with the previous three years and direct or indirect third-party involvement in half of reported incidents. The nonperforming loan coverage approach will be extended gradually and will account for individual institutions’ circumstances. Certain activities, particularly the acquisition and management of nonperforming loans and factoring of claims against public authorities, may be excluded case by case to avoid distortions and regulatory arbitrage.

Original source