Bank of England (BoE) to explore AI techniques in detecting criminal activities within retail payments data.
This partnership aimed at enhancing real-time detection mechanisms for identifying financial crime patterns. The joint project, named after the British inventor and suffragette Hertha Ayrton, focused on utilizing transaction analytics to spot complex criminal activity.
Results of Project Hertha
The outcomes revealed that AI techniques could significantly boost the identification of illicit accounts by an additional 12%. Furthermore, payment system analytics contributed substantially in uncovering novel financial crime patterns, marking a 26% increase in detection rates.
During testing, artificial intelligence generated a dataset simulating transactions for over 1.8 million bank accounts and more than 308 million transactions within a single jurisdiction’s retail payment system. The AI model was trained to mimic realistic transaction behaviors, creating a comprehensive ecosystem.
Challenges and Insights
While the results are promising, the project also highlighted limitations associated with practical implementation and systemic analysis effectiveness. BoE and BIS acknowledged that introducing such solutions would face significant challenges including regulatory and legal considerations.
The study indicated that payment system analytics were particularly effective in detecting intricate schemes involving multiple accounts across different banks and payment service providers (PSPs). For certain cases, detection accuracy doubled. However, for optimal performance, algorithms should be trained using verified past examples and require supervised learning.
As BIS transitions and prepares for leadership changes, the Innovation Hub is continuing its focus on fintech developments and central bank initiatives aimed at optimizing the global financial system’s operations. The collaboration with BoE underscores the ongoing efforts to fortify payment systems against financial crimes.