Swift, in collaboration with 13 global financial institutions, has recently conducted experiments that highlight the potential of AI and secure cross-border data sharing to reduce fraud in international payments.

Fraud Insights Sharing Through Privacy-Enhancing Technologies

The experiments employed privacy-enhancing technologies (PETs) to facilitate secure collaboration on fraud insights among participating institutions. Notable participants include ANZ, BNY, Intesa Sanpaolo, and technology partner Google Cloud.

One use case involved real-time verification of suspicious accounts using PETs, potentially speeding up the detection of complex international financial crimes and preventing fraudulent transactions. Another scenario utilized federated learning combined with PETs to identify abnormal activities without sharing customer data. This model, trained on synthetic data from ten million artificial transactions across all participants, demonstrated twice the effectiveness in recognizing known fraud cases compared to a model based solely on one institution’s dataset.

Future Plans for Expanding Participation

Following these successful experiments, Swift plans to broaden participation and move into a second phase of testing. This will involve using real transaction data to showcase how the technologies can impact actual fraud scenarios. Swift representatives emphasized that the company aims to facilitate secure intelligence sharing across borders to reduce annual fraud losses and enhance fraud prevention.

Swift has been actively exploring AI applications in cross-border payments, currently having more than 50 use cases across proof of concept, pilot projects, and live implementations.