Swift’s AI Breakthrough Against Financial Fraud
Swift, the global financial messaging network, has made a significant advancement in combating financial fraud by successfully conducting trials that doubled real-time fraud detection rates. This was achieved through the application of artificial intelligence combined with secure data sharing across a consortium of 13 leading banks, including ANZ, BNY Mellon, and Intesa Sanpaolo. During these trials, AI systems analyzed over 10 million transactions to identify suspicious activity more swiftly and accurately than traditional methods.
The Technology: PETs and Federated Learning
The breakthrough centers on two cutting-edge technologies: Privacy-Enhancing Technologies (PETs) and Federated Learning. PETs enable participating financial institutions to share insights about fraudulent activities securely without exposing sensitive customer data. This method allows verification of suspicious accounts and transaction patterns in real time while preserving privacy.
Federated Learning complements this by allowing AI models to be trained locally within each bank’s own data environment. Instead of pooling data centrally—which raises privacy and security concerns—the AI aggregates learning from multiple banks without transferring sensitive information. This decentralized approach results in more accurate identification of anomalies across different jurisdictions and banking systems.
Impact and Future Implications
The application of these technologies has delivered tangible benefits by drastically reducing the detection time for fraudulent transactions from days or hours to mere minutes, thereby potentially saving billions in losses globally. Key players in the trial included banking giants ANZ, BNY Mellon, Intesa Sanpaolo, and technology leader Google Cloud, which provided the infrastructure to support these AI initiatives.
Swift plans to expand this approach by increasing the number of participating institutions and transitioning trials from simulated to real transaction data. These steps are part of a broader AI strategy aimed at improving security and efficiency in cross-border payments, positioning Swift as a pioneer in secure and collaborative AI deployment in the financial sector.
Why This Matters for Banking AI
This development exemplifies a new paradigm in deploying AI for banking security—enabling secure collaboration among competing institutions to combat fraud at scale. By combining PETs and Federated Learning, Swift offers an advanced and privacy-conscious method for detecting financial crime without compromising customer data. The result is an elevated standard for AI adoption in finance, demonstrating the potential for technology to transform security protocols and operational effectiveness in a highly regulated industry.
