SWIFT Tests AI & Data Tools to Tackle Cross-Border Payment Fraud #
Swift has conducted pilot experiments with 13 global financial institutions using privacy-enhancing technologies (PETs) and federated learning to detect fraud in international payments. These tools let banks share insights on suspicious activity without exposing sensitive customer data, aiming to speed up detection of cross-border crime
One experiment, using synthetic data from 10 million artificial transactions, showed that models trained across institutions were twice as effective at spotting known fraud patterns than those relying on data from a single bank. Swift says the trials also included real-time verification of flagged accounts and that phases with actual transaction data are planned
Rachel Levi, Swift’s Head of AI, emphasized that a unified, cooperative defense offers greater protection than isolated efforts, particularly given that fraud costs financial services firms an estimated US$485 billion in 2023. The project aims to move into live environments with more participating banks and refined tools.