Europe’s banks gaining ground

Europe’s banks are making a regulatory bet on artificial intelligence: the same rules that slowed them down may now help them scale it. (bankingsupervision.europa.eu) The European Central Bank said on November 20, 2025 that AI use cases at supervised banks rose sharply between 2023 and 2024, based on reporting from 107 significant institutions in 2023 and 110 in 2024. Its 2025 workshops with 13 banks focused on credit scoring, fraud detection, governance and compliance. (bankingsupervision.europa.eu) The European Banking Authority said in its November 2024 Risk Assessment Report that AI adoption had “consolidated significantly” across the European Union and European Economic Area banking sector over the previous five years. Its 2024 survey showed the most common uses included profiling or clustering clients and transactions, customer support, anti-money laundering controls and fraud detection. (eba.europa.eu) The rulebook is now concrete. The General Data Protection Regulation has applied since May 25, 2018, the European Union’s AI Act entered into force on August 1, 2024, and the Digital Operational Resilience Act started applying on January 17, 2025. (commission.europa.eu 1) (commission.europa.eu 2) (eiopa.europa.eu) For banks, the AI Act lands hardest on lending. The European Banking Authority said in November 2025 that AI systems used to evaluate a person’s creditworthiness or set a credit score are classified as “high-risk,” triggering extra safeguards. (eba.europa.eu) That does not mean banks are starting from zero. The same European Banking Authority paper said its 2025 mapping exercise found “no significant contradictions” between the AI Act and existing European Union banking and payments law, describing the new regime as complementary to rules banks already follow on governance, risk and controls. (eba.europa.eu) Supervisors are also pushing banks toward the plumbing that large-scale AI needs. The European Central Bank said banks in its 2025 workshops had to update governance and compliance frameworks, while its revised internal-model guide in July 2025 added clarifications on the use of machine learning techniques under bank supervision. (bankingsupervision.europa.eu 1) (bankingsupervision.europa.eu 2) The counterargument is simple: heavy regulation can still slow deployment. The European Commission was still preparing practical guidance on how to classify high-risk systems in 2025, with guidelines due by February 2, 2026, leaving banks and vendors to work through open questions on who counts as a provider, who counts as a deployer and which systems fall into the strictest bucket. (eba.europa.eu) (plmj.com) European supervisors are not treating that uncertainty as a reason to wait. Their recent message has been that banks already using AI in credit scoring and fraud detection should tighten data quality, documentation and model oversight now, before generative AI spreads further into core banking decisions. (bankingsupervision.europa.eu) The result is a quieter race than the one in Silicon Valley: fewer public demos, more inventories, controls and audit trails. In banking, that back-office work often decides which systems reach production. (eba.europa.eu) (bankingsupervision.europa.eu)