BARCELONA, SPAIN – MARCH 1: Display of the MasterCard company logo on the stand during Mobile World Congress on March 1, 2017 in Barcelona, Spain. (Photo by Joan Cros Garcia/Corbis, Getty Images)
Joan Cross Garcia – Corbis | Corbis News | Getty Images
Ajay Bhalla, president of Mastercard’s cyber intelligence business, told CNBC that the new AI solution is powered by Mastercard’s proprietary recurrent neural network, built from the ground up by the company’s cybersecurity and anti-fraud teams. (The core part of generative AI).
“We’re basically using a transformer model that helps us get the power of generative AI,” Bhalla told CNBC in an exclusive interview earlier this week. “It’s all built in-house and we pull all kinds of data from the ecosystem. Because of the nature of the business we’re in, we look at all the transactional data that comes from the ecosystem.”
In some cases, Mastercard relies on open source “whenever necessary,” Barra added, but the “vast majority” of the technology is created in-house.
Mastercard’s proprietary algorithms are trained on data from approximately 125 billion transactions that pass through the company’s card network annually.
This data is not the words that are valued in large-scale language models such as OpenAI’s GPT-4 or Google’s Gemini, but rather helps AI understand relationships between merchants and where fraudulent transactions are taking place. Mastercard said it will help make predictions.
Instead of text input, Mastercard’s algorithms use the cardholder’s history of merchant visits as prompts to determine whether the businesses involved in the transaction are places the customer is likely to go.
The algorithm then generates a path through Mastercard’s network (like a heat-sensing radar) to find the answer in the form of a score.
A higher score is one that follows a pattern of the normal kind of behavior expected of a cardholder, while a lower score is one that deviates from that pattern.
According to Mastercard, this entire process takes just 50 milliseconds.
Barra said Mastercard’s new transaction decision-making technology could improve fraud detection rates by an average of 20% for financial institutions. However, Barra added that in some cases, the model has improved fraud detection rates by as much as 300%.
Mastercard says it has invested more than $7 billion in cybersecurity and AI technology over the past five years.
This includes a number of acquisitions, including a March 2023 deal to acquire Swedish cybersecurity company Baffinbay Networks.
Rival Visa has made its own investments in AI, including a $100 million venture fund for generative AI startups it launched in October 2023.
Although still in its early stages, Mastercard expects its algorithms to help banks eliminate much of the cost they typically spend evaluating fraudulent transactions, leading to savings of up to 20%.
According to Bhalla, the true potential of Mastercard’s technology lies in its ability to identify patterns and trends in fraud in order to predict future types of fraud that are currently unknown within the payments ecosystem. .
“The benefit of the Mastercard ecosystem is that we can see data from these transactions for all of our customers around the world,” he said. “What it does is it helps us really see fraud and patterns across the ecosystem around the world.”
Several companies in the payments and digital banking space have recently discovered that AI
Major changes to the product. PayPal announced a new AI-based product and one-click checkout feature last week.
clock: Mastercard launches Shopping Muse, an AI-powered personal retail assistant
