The BIS Innovation Hub’s London Centre, in collaboration with the Bank of England, has published the findings from Project Hertha, a study exploring how AI can help detect financial crime in real-time retail payment systems, while safeguarding user privacy.
Building on the earlier work of Project Aurora, Project Hertha focused on a key challenge in modern payments: how to identify complex criminal activity that is often spread across multiple bank accounts and financial institutions.
Because payment systems process transactions across many different players, they offer a rare opportunity to spot patterns that would otherwise remain hidden.
Project Hertha tested the use of advanced AI to analyse payment system data and uncover coordinated or unusual behaviour that may indicate financial crime.
To carry out this research responsibly, the team developed a state-of-the-art synthetic dataset containing 1.8 million simulated bank accounts and 308 million transactions.
This dataset was created using AI models that mimic real-world payment patterns, without using any personal customer information.
The results are promising.
By tapping into system-wide transaction analytics, banks and payment service providers were able to detect 12% more illicit accounts than they would through standard approaches.
When it came to identifying new or previously unseen financial crime patterns, the system achieved a 26% improvement in detection rates.
This suggests that AI-powered analytics could serve as a valuable complement to existing tools in the fight against financial crime.
At the same time, the project acknowledged that analytics alone are not a complete solution.
Rolling out similar systems in the real world would raise a number of legal, regulatory, and operational questions that need to be carefully considered.
These aspects were beyond the scope of the current research.
Project Hertha also emphasised the importance of using high-quality training data, creating a strong feedback loop to refine the AI models, and ensuring that the technology is transparent and explainable.
These elements are essential for building systems that are both effective and trustworthy.
The project is named in honour of Hertha Ayrton, the pioneering British scientist, inventor, and suffragette.
In 1904, Ayrton became the first woman to present a paper at the Royal Society, and in 1906 she received the prestigious Hughes Medal for her contributions to electrical science.
Featured image credit: Edited by Fintech News Switzerland, based on image by Alicja Ziajowska via Unsplash
