Visa bars $25 billion in fraud with AI

Visa Advanced Authorization (VAA) using artificial intelligence (AI) helped financial institutions prevent an estimated $25 billion in annual fraud making the global payment ecosystem safer for retailers and consumers, new analysis from Visa shows.

VAA is a comprehensive risk management tool that monitors and evaluates transaction authorizations on the Visa global payment network, VisaNet, in real time to help financial institutions promptly identify and respond to emerging fraud patterns and trends.

In 2018, Visa processed more than 127 billion transactions between merchants and financial institutions on VisaNet, and used AI to analyze all of these transactions — at about one millisecond each — so financial institutions can approve legitimate purchases while quickly identifying and preventing fraudulent transactions.

“One of the toughest challenges in payments is separating good transactions made by account holders from bad ones attempted by fraudsters without adding friction to the process,” said Melissa McSherry, SVP and global head of data, risk and identity products and solutions at Visa.

“By striking the right balance between human expertise and technology innovation, we continue to evolve our capabilities as new AI breakthroughs expand the realm of what’s possible,” McSherry said.

More than 8,000 financial institutions in 129 countries use VAA, a layer of fraud prevention that can help drive down risk and fraud for financial institutions and retailers, and help reduce friction due to false declines for payment account holders.

In 1993, Visa pioneered using neural networks modeled after the human brain to power its AI platform to identify possible fraud. This delivers faster and deeper insights through previously unknown correlations.

Delivered through VAA, retailers and financial institutions benefit from machine learning models used for real-time examination of each transaction for indicators of fraud – looking at activities, patterns and more than 500 risk attributes.