WASHINGTON — Visa and Mastercard are advancing separate but complementary artificial intelligence initiatives that signal how major card networks expect AI to reshape shopping, payments and fraud prevention.
The announcements highlight a growing push across the payments industry to integrate so-called “agentic” and generative AI tools directly into transaction systems, allowing machines to both analyze and, in some cases, initiate purchases on behalf of consumers.
Mastercard said it is developing a new large-scale AI “foundation model” trained on billions of anonymized transactions, with plans to expand to hundreds of billions of data points across its network.
‘Large Tabular Model’
Unlike traditional large language models used in chatbots, Mastercard said its system is a “large tabular model” designed to analyze structured payments data, including transaction histories, fraud patterns, merchant data and chargebacks.
Range of Services
The company said the model is intended to serve as an “insights engine” to enhance a wide range of services, including:
- Cybersecurity and fraud detection
- Loyalty and rewards programs
- Personalization tools
- Small business analytics
Mastercard said early testing shows the model can better distinguish legitimate purchases from suspicious activity, potentially reducing false fraud alerts on high-value but infrequent transactions.
Nvidia & Databricks Are Partners
The company is developing the system in collaboration with Nvidia and Databricks, leveraging high-performance computing to process large-scale transaction data.
Mastercard said it removes personal data from transactions used to train the model and is emphasizing privacy, governance and transparency as it expands the system’s capabilities.
Visa Launches Pilot
Visa, meanwhile, announced “Visa Agentic Ready,” a program designed to help banks test payments initiated by AI agents — software that can search, evaluate and complete purchases on behalf of consumers.
The program is launching in Europe with 21 issuing partners, including Barclays, HSBC UK and Banco Santander.
In the pilot phase, participating banks will run agent-initiated transactions in controlled, production-level environments alongside merchants to better understand how the payments behave and how to manage associated risks.
Visa said the effort is aimed at preparing financial institutions for a future in which AI-driven shopping tools become more common.
“As AI agents increasingly shape how people shop and buy, payments need to keep up,” said Mathieu Altwegg, head of product and solutions at Visa Europe, in a statement.
Security & Infrastructure
Visa said it is relying on existing payment security technologies rather than building entirely new systems for AI-driven transactions, including:
Tokenization, which replaces card numbers with digital tokens to protect sensitive data
Biometric authentication, such as fingerprint or facial recognition
Risk scoring and spending controls, allowing consumers to pre-set limits on agent activity
Under the model, consumers authorize parameters in advance, while AI agents execute transactions within those boundaries.
The Shift
Together, the initiatives reflect a broader shift by major payment networks toward embedding AI deeper into commerce infrastructure — from predictive analytics and fraud detection to autonomous purchasing.
While Mastercard is focusing on using AI to interpret and predict transaction behavior, Visa is testing how AI systems might actively participate in commerce.
Both companies said their efforts are designed to increase speed, security and personalization in payments, as financial institutions prepare for a more automated, AI-driven marketplace.
