International payments network Swift is deploying technology to make its payments faster, safer and more efficient. 

Swift is investing in technology because as “more and more [payments] become instant or near real-time,” stopping fraud is essential, Johan Bryssinck, program head of AI/ML at Swift, said at the AI for Financial Services 2024 conference on June 12.

Once a fraudulent actor receives a payment, it is very difficult to retrieve, he added. 

Swift is deploying three types of technology to increase security: 

1. Data sharing 

Swift is enabling financial institutions to share data to track and stop money laundering efforts, Bryssinck said.  

After offering FIs in the United States data sharing capabilities, Swift observed a fivefold increase in cases that should be investigated, Bryssinck said, adding that in the U.K. this year, attempted fraudulent transactions worth 1.4 billion pounds ($1.7 trillion) were stopped. 

Data sharing also allows Swift to fine-tune its AI models with the bank data, Bryssinck said. 

2. Artificial intelligence 

With billions of payments and billions of dollars flowing through its network, Swift is looking to deploy AI to “detect anomalies” quickly, Bryssinck said. 

Last month, Swift launched Payments Control, an AI-driven solution for financial institutions to create an accurate picture of potential fraud activity, according to May 30 release. 

“By adding a logistic regression model over a rule-based system, we saw a 40% reduction in false positives,” Bryssinck said, adding that FIs can create their own risk scores to train and fine-tune the model to track suspicious activity. 

Swift is also looking to use AI to aid banks in sharing transactional information with each other anonymously, Bryssinck said. 

3. Synthetic data 

Swift is tapping AI services provider Mostly AI to develop synthetic data to help train its AI models, Tobias Hann, chief executive of Mostly AI, said during the event.  

“You can actually modify the data during the creation process [and] create more rare events in a large data set” through synthetic data to train AI models for specific scenarios, Hann said, adding that synthetic data can aid in consumer data security. 

Editor’s note: All amounts have been converted to U.S. dollars.  

Early-bird registration is now available for the inaugural Bank Automation Summit Europe 2024 in Frankfurt, Germany on Oct. 7-8! Discover the latest advancements in AI and automation in banking. Register now.