Bank-to-bank payment network Swift is collaborating with Google Cloud to create advanced anti-fraud technologies that leverage AI and federated learning across multiple jurisdictions.
In the first half of 2025, Swift plans to introduce a sandbox using synthetic data to prototype solutions for learning from historical payment fraud. This initiative will involve 12 global financial institutions, with Google Cloud serving as a strategic partner. It builds upon Swift’s existing Payment Controls Service (PCS) and follows a successful pilot project with financial institutions in Europe, North America, Asia, and the Middle East.
In this collaboration, Google Cloud, along with technology partners such as Rhino Health and Capgemini, is working to develop a secure, privacy-preserving system that enables financial institutions to combat fraud effectively. The strategy utilizes federated learning techniques coupled with privacy-enhancing technologies (PETs) to facilitate data sharing among institutions without exposing proprietary information.
Rhino Health is responsible for creating and delivering the core federated learning platform, while Capgemini will oversee the system’s implementation and integration.
Rachel Levi, head of artificial intelligence at Swift, comments, “Swift occupies a unique position in the financial industry as a trusted and collaborative network essential to the global economy. We are well-placed to spearhead industry-wide efforts to combat fraud.”
She adds that this initiative seeks to validate whether federated learning technology can allow financial institutions to anticipate and counteract fraud by sharing insights, which will ultimately enhance the cross-border payments experience for customers.
Federated learning provides a robust solution for collaborative AI model training without compromising data privacy and confidentiality. Instead of requiring the pooling of sensitive data, model training occurs within individual financial institutions on decentralized datasets.
Only the insights gained from this training, and not the raw data itself, will be transmitted back to a central server managed by Swift for aggregation.
Andrea Gallego, managing director of global go-to-market incubation at Google Cloud, states, “Our partnership with Swift showcases the transformative potential of federated learning and confidential computing. By promoting secure collaboration and knowledge sharing without sacrificing data privacy, we are helping to build a safer and more resilient financial ecosystem for all.”