Bank for International Settlements Innovation Hub reports Project Hertha results showing payment system analytics can identify 12% more illicit accounts

The Bank for International Settlements Innovation Hub has published findings from Project Hertha, a joint project between its London Centre and the Bank of England, testing whether transaction analytics using a minimal set of data points can help detect financial crime patterns in real-time retail payment systems while supporting user privacy. The experiments indicate that payment system analytics can serve as a supplementary tool for banks and payment service providers (PSPs) to identify suspicious activity, particularly where criminals operate across complex networks spanning multiple institutions. In testing modern artificial intelligence techniques on payment-system-level data, the project found that incorporating payment system analytics helped banks and PSPs identify 12% more illicit accounts than they would otherwise have found, with a 26% improvement when attempting to detect previously unseen behaviours. The work used a simulated synthetic dataset developed for the project, covering 1.8 million bank accounts and 308 million transactions, generated with an AI model designed to produce realistic transaction patterns without using real customer data. The report also flags limits to the approach and notes that deploying similar solutions would raise practical, legal and regulatory issues that were outside the project’s scope, while highlighting the importance of labelled training data, robust feedback loops and explainable AI to maximise effectiveness.