Google Cloud and Swift Partner on AI Model Training

Google Cloud and Swift are working together to develop a secure, privacy-preserving solution for financial institutions that use federated learning techniques.

Federated learning enables collaborative artificial intelligence (AI) model training without compromising privacy and confidentiality, Google Cloud said in a Tuesday (Dec. 10) blog post.

Swift plans to roll out a sandbox with synthetic data in the first half of 2025, working with Google Cloud and 12 global financial institutions, according to the post.

“This exploration will help the community validate whether federated learning technology can help financial institutions stay one step ahead of bad actors through sharing of fraud labels, and in turn enabling them to provide an enhanced cross-border payments experience to their customers,” Rachel Levi, head of AI at Swift, said in the post.

Technology partners including Rhino Health and Capgemini will also collaborate on this project, according to the post. Rhino Health will develop and deliver the core federated learning platform, while Capgemini will manage the solution’s implementation and integration.

The solution’s components include a federated learning server that orchestrates the clients’ collaboration, a federated client that executes tasks and submits results for secure aggregation, and bank-specific encrypted data that remains encrypted through the entire process, the post said.

It also includes a global fraud-based model, secure aggregation and a global-anomaly-detection trained model, per the post.

“Our collaboration with Swift exemplifies the transformative potential of federated learning and confidential computing,” Andrea Gallego, managing director, global go-to-market incubation at Google Cloud, said in the post. “By enabling secure collaboration and knowledge sharing without compromising data privacy, we are fostering a safer and more resilient financial ecosystem for everyone.”

Swift said in October that it was bolstering its AI-enhanced fraud detection capabilities and will begin offering this service to the payments sector beginning in January.

The new capability expands on Swift’s Payment Controls Service by using pseudonymized data from the billions of transactions moving across the Swift network to spot suspicious transactions, allowing for real-time response.

Swift said at the time that the effort is part of its wider partnership with its network of more than 11,500 banks and financial institutions to determine how AI “can solve cross-industry challenges.”