Swift banks on AI to fight global fraud

Swift, the world’s leading bank messaging network, has announced successful results from trials combining artificial intelligence and privacy-enhancing technologies (PETs) to combat financial fraud. The cooperative worked with 13 major banks and technology partners, including Google Cloud, testing new methods on ten million synthetic transactions. The trials demonstrated that fraud detection rates doubled compared to existing models.

Real-Time Cross-Border Intelligence Sharing

One innovative use case enabled institutions to share fraud insights across borders in real time using PETs. This approach allowed participants to verify intelligence on suspicious accounts without exposing customer data. Such a framework could significantly reduce the time needed to identify and stop international financial crime networks, helping prevent fraudulent transfers from completing.

Another test employed federated learning, an AI technique that trains models locally within each institution while protecting sensitive information. By leveraging combined knowledge from multiple participants, the model proved twice as effective at identifying fraudulent activity compared to models using only single datasets.

Industry Participation and Impact

The pilot program included leading financial institutions such as ANZ, BNY, and Intesa Sanpaolo. According to Swift’s head of AI, Rachel Levi, this secure intelligence sharing could reduce industry losses by billions of dollars annually. She emphasized that the approach allows fraud to be stopped within minutes rather than hours or days.

Future Expansion and Broader Initiatives

Following these successful trials, Swift will expand participation and move into a second testing phase. This next stage will use real-world data to demonstrate the technology’s effectiveness in active banking environments. The cooperative has been investing heavily in AI, with more than 50 ongoing use cases across various development stages.

Earlier this year, Swift also launched an AI-powered Payments Controls Service designed to help smaller financial institutions identify suspicious transactions more effectively. Together, these initiatives represent a significant push toward building smarter, faster, and more collaborative fraud prevention systems across global payment networks. This development marks a major advancement in financial security technology, potentially transforming how banks worldwide combat increasingly sophisticated financial crimes.

AI fraud detectionbanking technologyCross-border paymentsfederated learningFinancial CrimeGoogle Cloudprivacy-enhancing technologiesSwift