The Dutch Banking Association (NVB) is calling on financial institutions to expand their use of machine learning models to monitor customer transactions, arguing that modern technology can significantly improve the accuracy and efficiency of detecting financial crime.
In a newly published white paper, the NVB expressed that traditional rule-based systems which rely heavily on fixed transaction-amount thresholds and rigid parameters generate hundreds of thousands of suspicious activity reports every year many of which end up being false alerts.
Unlike rigid traditional setups, machine learning algorithms make predictions by analyzing patterns from past data and cross-analyzing information across multiple behavioral dimensions simultaneously.
This approach offers key advantages by allowing AI models to score transaction alerts and automatically clear low-risk activity which significantly reduces unnecessary flags and eases customer friction.
Additionally, advanced algorithms can identify subtle, multi-layered patterns that traditional systems usually miss while enabling banks to conduct smarter customer reviews based on significant account changes rather than fixed calendar intervals.
The NVB outlined two main structural solutions banks can deploy, starting with general models that serve as broader frameworks featuring various parameters representing different risk levels across the financial crime landscape. Alternatively, institutions can deploy focus models which are smaller specialized models tailored specifically to detect individual financial crime risks.
See also: Bank of America Vice President Killed in Random Times Square Stabbing
To ensure safe adoption, the NVB emphasized that machine learning models must remain transparent, interpretable, and fully explainable to regulators, analysts, and compliance officers. The lobby group recommended specific techniques such as adversarial debiasing and subgroup calibration to ensure algorithms remain fair and prevent bias against specific customer demographics based on attributes like age or gender.
Despite the efficiency gains offered by machine learning, the NVB stressed that technology will never fully replace human judgment. Financial institutions remain legally accountable for all decisions made by these systems making active human oversight essential to monitor and control AI tools.
Human supervision will become even more critical under upcoming European Union anti-money laundering regulations taking effect in 2027 which mandate strict human checks to guarantee a model’s accuracy and appropriateness before automated decisions impact customers.
