Visa is deploying AI to fight fraud on instant payment networks as businesses and consumers ramp up the use of the payment channels.

The payments giant developed the solution in-house, James Mirfin, senior vice president and global head of risk and identity solutions, told Bank Automation News.  

“With instant payments, you cannot make a mistake. Once the payment is out the door, it’s really difficult to get it back,” Mirfin said. “Also, new payments rails are more prone to fraudulent activities until FIs figure out how fraudsters are exploiting the system.” 

Fraudulent transactions are more difficult with traditional card payment methods because fraudsters must find multiple pieces of information and overcome more sturdy fraud fighting mechanisms that have been developed over years, he said. 

“In a real-time payment, you really only have one opportunity to challenge the consumer,” which makes it easier for the fraudster, Mirfin said. 

The AI-driven instant payments model has been deployed in Argentina and the United Kingdom and has an efficacy rate of 50% to 70%, Mirfin said. Visa plans to add its AI model for instant payments to other networks this year. 

To reduce fraud on instant payment networks, Visa’s AI gives a risk score to the transaction based on: 

• Whether it is a new type of payment to an unusual merchant; 

• Payment amount; and 

• Personal information like an account holder name, email and phone number. 

Fraud fighting models must calculate the risk quickly, which requires high computing power and a great deal of data to conduct training on, Mirfin said. Companies are still building data sets to train AI models since instant payment networks are relatively new. 

“These models are scoring transactions in a single millisecond. … For context the blink of an eye is about 100 milliseconds,” Mirfin said.  

The payment processor aims to develop ultra-low latency and ultra-high frequency AI models to stop fraud while offering a seamless experience to consumers. 

“It’s about finding the right balance between predictive and generative AI,” Mirfin said. The model must be able to process advanced authorization for transactions and flag them quickly and accurately if they seem fraudulent. 

Visa recently acquired Featurespace, an AI-driven fraud-fighting company, for an undisclosed amount “to provide an expanded set of fraud prevention tools to our clients and protect consumers in real time across various payment methods,” Chief Executive Ryan McInerney said during the company’s first fiscal quarter earnings call on Jan. 30. 

The credit card giant aims to expand its fraud-fighting capabilities to 10 new real-time payments network in 2025, McInerney said on the call. 

Register here for Bank Automation Summit 2025, taking place March 3-4 in Nashville, Tenn. View the full event agenda here.