BARCELONA, SPAIN – MARCH 1: A view of the MasterCard company logo on their stand during the Mobile World Congress on March 1, 2017 in Barcelona, Spain. (Photo by Joan Cros Garcia/Corbis via Getty Images)
Joan Cros Garcia – Corbis | Corbis news | Getty Images
Payment giant MasterCard says it has developed its own proprietary generative artificial intelligence model to help thousands of banks in its network detect and root out fraudulent transactions.
The company exclusively told CNBC that its new advanced AI model, Decision Intelligence Pro, will allow banks to better assess suspicious transactions on its network in real time and determine whether they are legitimate or not.
Ajay Bhalla, president of Mastercard’s cyber and intelligence business unit, told CNBC that the new AI solution is a proprietary recurrent neural network – a core part of generative AI – powered by Mastercard’s cybersecurity and The company’s anti-fraud team was built from the ground up.
“We use the transformer models, which essentially help harness the power of generative AI,” Bhalla said in an exclusive interview with CNBC earlier this week. “It’s all built in-house, we have all kinds of data from the ecosystem. Due to the nature of the business we are in, we see all the transactional data that comes to us from the ecosystem.”
In some cases, Mastercard relies on open source “whenever necessary,” but the “majority” of the technology is developed internally, Bhalla added.
Mastercard’s proprietary algorithm is trained on data from the approximately 125 billion transactions processed annually across the company’s card network.
The data helps AI understand relationships between traders – not words, as is the case with large language models like GPT-4 and OpenAI Googles Gemini – and predict where fraudulent transactions are taking place, Mastercard said.
Heat detection fraud patterns
Instead of text input, Mastercard’s algorithm uses a cardholder’s merchant visit history as an indication to determine whether the business involved in a transaction is a location the customer would likely go to.
The algorithm then generates paths through Mastercard’s network – a type of heat-sensitive radar – to find the answer in the form of a score.
A higher score would be one that follows the pattern of usual behavior expected of the cardholder, and a lower score is outside that pattern.
According to Mastercard, this process only takes 50 milliseconds.
According to Bhalla, Mastercard’s new transaction decision technology can help financial institutions improve their fraud detection rates by an average of 20%. However, in some cases, the model has resulted in up to 300% improvement in fraud detection rates, Bhalla added.
Mastercard says it has invested more than $7 billion in cybersecurity and AI technologies over the past five years.
This includes a number of acquisitions, including the March 2023 deal to buy Swedish cybersecurity company Baffin Bay Networks.
Competitors Visas has made its own investments in AI, including a $100 million venture fund for generative AI startups that the company launched in October 2023.
Although it’s still early, Mastercard believes its algorithm will allow banks to save up to 20% by eliminating much of the cost they would normally incur in assessing illegal transactions.
The true potential of Mastercard technology, according to Bhalla, lies in its ability to identify fraudulent patterns and trends to predict future types of fraud that are currently unknown in the payments ecosystem.
“The beauty of the Mastercard ecosystem is that from these transactions we see data from all of our customers worldwide,” he said. “This helps us actually detect fraud and patterns across the ecosystem globally.”
Several companies in the payments and digital banking space have recently stated that AI will lead to this
big changes to their products. PayPal Last week, the company announced new AI-based products as well as a one-click checkout feature.
REGARD: Mastercard introduces Shopping Muse, an AI-powered personal retail assistant
Source link
2024-02-01 06:00:31
www.cnbc.com
