Recent research from the BIS Innovation Hub emphasizes that leveraging payments data, privacy-enhancing technologies, AI, and enhanced cooperation for a behavioral-based analysis approach is significantly more effective in detecting money laundering networks than traditional rules-based methods.
The Financial Action Task Force notes that the majority of substantial money laundering schemes are cross-border and span various business sectors. However, financial institutions often struggle to identify potential suspicious networks and transactions due to fragmented data and disparate systems.
A study by LexisNexis Risk Solutions reveals the financial strain AML compliance imposes on financial institutions, with costs rising approximately $60 billion from 2020 to 2022, totaling around $274 billion.
The BIS Innovation Hub’s Nordic Centre has been exploring innovative solutions and conducted a proof of concept known as Project Aurora in collaboration with Lucinity, an Icelandic AI software-as-a-service company. This proof of concept utilized a comprehensive synthetic dataset that emulates real-world domestic and international payment transactions while incorporating privacy-enhancing technologies to protect sensitive information through encrypted data.
Algorithms were subsequently trained on this synthetic dataset to identify various patterns, known as "typologies," associated with money laundering across different institutions and countries. The project examined different perspectives on the synthetic data to create a range of monitoring scenarios, including siloed, national, and cross-border frameworks. Various collaborative analysis approaches, including centralized, decentralized, and hybrid models at both national and international levels, were also assessed.
The findings underscore the effectiveness of advanced analytics and technologies that utilize a behavioral-based analysis approach, which focuses on understanding the relationships between individuals and businesses and identifying deviations from typical behavior. This research indicates that such methodologies outperform traditional rules-based approaches, which are hindered by their isolated nature.