European Central Bank Banking Supervision published a Supervision Newsletter article on banks’ use of artificial intelligence (AI) for credit scoring and fraud detection, based on its annual data collection, on-site inspections and a series of workshops with 13 banks. The analysis finds that AI use cases among significant institutions increased between 2023 and 2024 and summarises observed practices and shortcomings relevant to microprudential supervision. Across the sampled banks, decision tree-based models were most commonly used for both credit scoring and fraud detection, while neural networks were mainly used for fraud detection. While banks reported efficiency and performance benefits, quantifying realised financial gains remained difficult, echoing on-site inspection findings on tracking the financial impact of digital initiatives. Governance responses included embedding AI into existing risk frameworks, with about half of the sample introducing dedicated AI policies or committees; banks also reported preparations for the EU AI Act, including self-assessments to identify high-risk use cases and mapping AI use across the organisation, alongside emerging roles such as Chief AI Officers and efforts to strengthen second and third lines of defence. On risk management, banks typically used explainability tools and centralised dashboards, often developed models in house but hosted them in cloud environments, and relied on human validation and oversight for higher-risk decisions; none of the sampled banks permitted self-learning after deployment. The article highlights growing attention to third-party dependencies, data privacy, operational resilience and compliance, with ECB Banking Supervision prioritising deficiencies in cybersecurity and third-party risk management capabilities; it also notes residual “black box” concerns, divergent interpretations of explainability, and limited effective application of data governance standards, reinforcing the need to continue follow-up work on risk data aggregation and risk reporting. AI remains a supervisory priority, with the supervisory priorities for 2026-2028 set to continue monitoring banks’ AI strategies, governance and risk management. ECB Banking Supervision plans continued industry engagement, consultation with external experts and further development of internal capabilities to identify, monitor and mitigate emerging risks.
European Central Bank - Banking Supervision 2025-11-20
European Central Bank Banking Supervision reports rising AI use for credit scoring and fraud detection and sets supervisory focus on governance and operational resilience
The European Central Bank Banking Supervision published a newsletter on banks' AI use for credit scoring and fraud detection, noting increased use among significant institutions between 2023 and 2024. The analysis identifies common practices, governance responses, and challenges, including difficulties in quantifying financial gains and concerns over third-party dependencies and data privacy. AI remains a supervisory priority, with plans for continued monitoring and industry engagement to address emerging risks.