Visa to Test Advanced AI to Prevent Fraud

The platform, built in house and slated to be launched later this year, is an example of the broader financial-services industry trend of using AI to detect patterns in transactions that could signal criminal behavior. The platform is cloud-based, meaning that Visa’s researchers and engineers can access it online from anywhere.

“One of the transformative technologies of this era is going to be AI,” said Rajat Taneja, executive vice president of technology and operations for Visa, the largest U.S. card network by cards in circulation and transactions. “There is a perfect combination right now of computing resources, algorithms, data and people that’s allowing this incredible innovation,” he added.

The banking industry is expected to be the second biggest spender on AI systems this year, behind retail, according to market-research firm International Data Corp. Banks are on track to spend $5.3 billion on AI in 2019, growing to $12.4 billion in 2023, on such initiatives as fraud analysis, according to IDC.

Visa said it has spent about $500 million over the past five years on AI and data-infrastructure projects.

The new platform is expected to test algorithms that use an advanced form of AI called deep learning, a technique that has the potential to identify more complex patterns than traditional machine-learning algorithms.

“It’s a massive breakthrough for us,” Mr. Taneja said.

Visa currently uses machine-learning algorithms to sift through data to identify anomalies, an effort that prevents billions of dollars in fraudulent transactions annually, Mr. Taneja said. One such Visa fraud-detection system, Advanced Authorization, prevented about $25 billion in fraud in the year ended April 30, according to the company.

But the current models have limitations. Researchers must know the signals that might indicate fraud—such as a purchase taking place at an unusual time of day—and write the rules to tell the model what to do when it identifies suspicious activity. Criminal activity sometimes slips by unnoticed because hackers are getting more sophisticated at evading the warning signs that current machine-learning models are trying to detect.

Deep-learning models can automatically identify more complex patterns by themselves. For example, if a customer uses his or her card in another country for the first time, deep-learning algorithms will be able to tell, with more accuracy and fewer false positives than traditional machine learning, whether it’s a legitimate transaction. The algorithms will be able to take into account previous transactions at airlines and hotels, as long as they are made with Visa cards.

The development of the AI platform comes as more consumer credit-card information is leaked in high-profile breaches. Americans reported losing $1.48 billion to fraud including identity theft in 2018, up 38% from 2017, according to the Federal Trade Commission’s analysis of more than 1.4 million fraud reports.

Scientists at Visa’s 56-person research team built the algorithm-testing system, which is akin to an internal cloud that runs on the company’s own data centers instead of using public-cloud services such as those from

Amazon.com
Inc.,

said Hao Yang, vice president of Visa Research.

Developers using the platform will be able to access a secure data set composed of Visa’s real-time card transactions in such a way that they can test algorithms on a subset of the data before deploying it widely, without disrupting the normal course of global card traffic.

In the past, the research team had to test algorithms on historical data and then hope they would work once they were deployed on real-time data, Mr. Yang said.

Write to Sara Castellanos at sara.castellanos@wsj.com