Bank-to-Bank messaging network Swift uses a combination of AI and cross-border data sharing to call out the results of tests that doubled real-time fraud detection on attempts that include 13 top tier banks and 10 million test transaction sons.
editorial
This content is selected, created and edited by the FineXtra editorial team based on its relevance and interest in the community.
This experiment used privacy-enhancing technology (PET) to enable agencies to safely share fraud insights across borders. In one use case, PETS allowed participants to verify intelligence in real time with suspicious accounts. This reduces the time it took to identify complex international financial crime networks and avoids fraudulent transactions that are being carried out.
In another use case, participants used a combination of pets and federated learning. This allows for the organization to “visit” each organization and locally train data, allowing them to cooperate without sharing customer information, identifying unusual transactions.
Trained with synthetic data from 10 million artificial transactions between participants, this model was twice as effective in identifying known fraud instances as models trained on a single institutional dataset.
Fiancé agencies involved in the exam include technology partners such as ANZ, BNY, Intesa Sanpaolo and Google Cloud.
Rachel Levi, head of AI at Swift, said: “The industry loses billions of dollars each year, but by allowing intelligence to safely share across borders, we can cut this number significantly and stop scams in minutes rather than hours or days.”
Following these successful experiments, Swift plans to expand participation before beginning the second phase of the test. This attempts to use real transactional data to demonstrate the impact of technology on real fraud.
Cooperatives are actively investigating the role AI can play in solving cross-border payment challenges, and currently have over 50 use cases beyond proof of concept, pilot and live usage.
Earlier this year, Swift launched the Ai-Enhanced Payments Controls Service. This helps small and medium-sized financial institutions more accurately flag suspicious transactions, allowing actions to be taken in real time.
