Belgium-based telecommunications company Swift has announced two AI experiments developed in collaboration with its member banks. The purpose of the experiments is to investigate the technology's potential in combating cross-border payments fraud and reducing associated costs for the financial industry.

In the first pilot, Swift aims to enhance its existing Payment Controls service by integrating an AI model designed to improve the detection of potential fraud through a detailed analysis of historical activity patterns on the Swift network. This enhancement will be refined with input from Payment Controls customers and will use live traffic data from these customers, ensuring the practical applicability of the findings. Separately, Swift has assembled a group of ten leading financial institutions to explore the use of advanced AI technology in analysing anonymised data from multiple sources.

This initiative aims to improve the global financial ecosystem's defences against fraud. The secure data collaboration and federated learning technologies being tested could lead to broader information-sharing practices in fraud detection, building on previous successes in cybersecurity threat assessment. The participating institutions include BNY Mellon, Deutsche Bank, DNB, HSBC, Intesa Sanpaolo, and Standard Bank.

They will leverage a secure infrastructure that allows financial institutions to exchange relevant information while maintaining strong privacy controls. Swift's AI anomaly detection model will analyse this dataset to identify potential fraud patterns more effectively. The cost of fraud According to Swift's press release, fraud inflicted a cost of USD 485 billion on the financial industry in 2023.

AI's role in reducing these costs could also support the G20’s objective of accelerating cross-border payments. Swift's central position in the financial ecosystem, combined with its extensive data network, positions it well to lead industry efforts against fraud. Swift has developed an AI governance framework with its community to ensure accuracy, explainability, fairness, security, and privacy in its AI applications.

These pilots are conducted under principles of responsible AI use and comply with emerging global standards, such as ISO 42001, the NIST AI Risk Management Framework, and the EU AI Act. .