Source: Iusletter

The OECD report “Artificial Intelligence in Italian Financial Markets”, published on 24 April 2026, documents rapid growth in AI use across Italian banking and insurance, based on an industry survey developed with Banca d’Italia, the European Commission and national authorities that collected 450 responses (49% response rate). Insurers are among the most advanced adopters, with 70% of insurance respondents reporting AI use versus 39% across all respondents, while banks report greater breadth and maturity with over 2,500 reported use cases and more deployments in production. Measured benefits include operational efficiency (three quarters of AI users), productivity gains (62%), internal process improvements (45%) and better front-office decision-making (49%), with applications spanning underwriting, pricing, claims management, anti-fraud, customer care and portfolio monitoring in insurance, and fraud prevention, risk management, document analysis, compliance support and process automation in banking.

The report also highlights governance and risk concerns that limit industrial adoption: 1,042 general-purpose AI (GPAI) use cases are in development or experimentation versus 713 in production, and 73% of GPAI users rely primarily on non-customized licensed models, raising responsibility issues for intermediaries that must still govern third-party models. Half of respondents cite human-in-the-loop as the main safeguard, while measures like incident management, input/output restrictions, backups, malfunction monitoring and data-integrity controls are less common; 46% have not implemented AI-specific cyber defenses. The analysis situates these findings within the EU regulatory shift driven by Regulation (EU) 2024/1689 (AI Act), which classifies certain financial systems as high-risk (including creditworthiness assessment and life/health pricing), and concludes that banks and insurers need documented mappings, risk classification, data governance, traceability, effective human supervision, impact assessments, incident reporting procedures and vendor management to turn experimentation into responsible, compliant adoption.