ThetaRay, a firm providing Cognitive AI financial crime compliance, has released a study on the future of anti-money laundering (AML) in Europe. The study warns that Europe’s anti-money laundering system is approaching structural failure, and that financial institutions will be unable to meet upcoming supervisory expectations without advanced AI-driven monitoring and customer-screening systems.
ThetaRay’s core service uses machine learning algorithms to detect suspicious activities, aiming to support safer cross-border payments.
The report, “Next-Generation AML Solutions: An Analysis of AI-Based Tools vis-à-vis the Reform of the European AML Institutional and Substantive Architecture,” examines how the EU’s sweeping AML reform package and the Artificial Intelligence Act will reshape compliance across the region.
Co-authored by Prof. Andrea Minto, an authority on EU financial regulation at Ca’ Foscari University of Venice and the University of Stavanger, and Yaron Hazan, ThetaRay’s Vice President of Regulatory Affairs, Advisory board member at the AI APAC Institute and former Head of Compliance at HSBC Israel, the study blends academic rigor with supervisory and operational expertise rarely found in a single publication.
The report finds that despite rising budgets and stronger enforcement, Europe’s AML framework continues to underperform:
- The FATF reports that 97% of 120 assessed countries show only low to moderate effectiveness in preventing money laundering and terrorist financing.
- FIUs across Europe report extremely low intelligence yield: the Netherlands identified <3.5% of 3.48 million 2024 reports as suspicious, while France’s Tracfin reports only ~5% actionable SARs.
- Germany’s FIU data shows that only 15% of SARs are investigated by law enforcement, with 95% of forwarded cases ending without prosecution.
- One operational risk study found that rules-based detection scenarios produced reporting in just 2% of cases.
According to the study, the global AML system suffers from persistent structural inefficiencies, high false-positive rates, and poor conversion from alerts to meaningful intelligence, the result of legacy rule-based systems that generate low-quality alerts, rely on siloed architectures, and lack the cross-border visibility required to detect modern networked financial crime.
“The data is clear: Europe’s AML system is no longer keeping pace with financial complexity,” says Hazan.
The report highlights how the EU’s two major regulatory initiatives, the AML Package and the Artificial Intelligence Act, together represent a consequential shift in AML expectations. The AML Package aims to strengthen due diligence obligations, expand governance requirements, and establish a new EU-level Anti-Money Laundering Authority (AMLA), harmonizing obligations across Member States. At the same time, the AI Act classifies transaction monitoring and sanctions screening as “high-risk” uses of AI, imposing strict requirements on transparency, human oversight, data governance, and model lifecycle management.
The authors also highlight vulnerabilities in correspondent banking and crypto-asset flows, where traditional rule engines struggle to detect hidden network behaviour across complex cross-border transaction chains. The study further identifies growing friction between the AML Regulation (AMLR) and GDPR data processing constraints, warning that without clearer guidance, institutions could face overlapping regulatory and legal risks.
Minto states;
“The AML Package and the AI Act make clear that the integration of AI into customer due diligence and AML monitoring is inevitable. Financial institutions must now prepare for a world in which technological capability and legal obligation are inseparable.”
The report ultimately calls for a fundamental shift from volume-driven alerting to intelligence-led detection, emphasizing hybrid human-AI oversight, robust data governance aligned with the AI Act, transparent and explainable models, and integrated customer and transaction screening workflows