Visa Uses AI to Strengthen Payment Security and Combat Organized Fraud

Account-to-account payments are expanding as banks and consumers seek faster ways to move money, but instant settlement leaves financial institutions less time to identify scams before funds are gone. Visa is combining its global network intelligence with technology from Featurespace, the British AI fraud-detection company it acquired in 2024, to help banks assess transaction risk without waiting for locally trained models or a consortium to build sufficient data.

Visa unveiled the enhanced A2A Protect on Sept. 1, 2026, adding a unified fraud score and real-time insights intended to stop suspicious transfers before money leaves customer accounts. The company said deployments have increased fraud detection by 75% within the first six months while reducing unnecessary fraud alerts by more than 40%. Global A2A transaction volumes are projected to exceed 5.8 trillion by 2028, up 160% from 2024, increasing the urgency for network-level defenses.

Visa is expanding beyond its role as a credit and debit card transaction network into value-added merchant services, focusing on agentic AI, stablecoins and digital assets. The shift could determine whether Visa can reduce its reliance on card-processing fees as payment methods rapidly evolve, while generating revenue from fraud prevention, data analytics and new settlement tools.

Visa executives recently said the company is advancing agentic commerce, in which AI agents can independently search, shop and make payments. It is also using AI to detect and block payment fraud such as romance scams while developing stablecoin applications. Reports did not provide investment figures, a formal launch date or projected new revenue, and Visa has yet to disclose a specific timetable.

Visa said in its Spring 2026 threat report that generative AI has sharply reduced the cost and technical barriers involved in creating phishing messages, impersonating identities and automating attacks. This allows criminal groups to target more consumers simultaneously. The focus of payment defenses is also shifting from preventing system breaches to identifying transactions made by real people who have been manipulated, creating new challenges for banks, merchants and card networks.

The latest report shows that both attackers and defenders are expanding their use of AI. Fraudsters use it to rapidly generate convincing content and test attacks at scale, while financial institutions deploy AI to strengthen real-time monitoring and anomaly detection. Visa warned that the key shift in Spring 2026 was not an increase in the value of individual fraud cases, but a simultaneous rise in the speed, volume and personalization of attacks, making them harder for traditional rules-based fraud controls to stop in time.

Agentic commerce allows software to search for, select and purchase products with a user’s authorization, making identity verification, payment authorization and risk controls critical. Visa launched Visa Intelligent Commerce on April 30, 2025, to provide payment APIs to AI platforms. Connect further integrates the connection process, lowering adoption barriers for merchants, agent developers and payment service providers.

Visa unveiled Intelligent Commerce Connect on April 8, 2026. Through a single integration with the Visa Acceptance Platform, businesses can initiate payments, tokenize credentials, set spending controls and perform authentication, with support for APIs from Visa and other card networks. Seven companies, including AWS, Mesh and Payabli, are piloting the service, with a broader rollout planned during 2026. Visa did not disclose pricing or an investment amount.

Financial scammers are shifting their focus from breaching systems to exploiting victims' trust, fear and sense of urgency. Even as banks and payment providers deploy AI to detect unusual transactions, criminals can still use impersonation and social engineering to persuade victims to authorize payments themselves. Visa says the trend matters because traditional models struggle to determine in real time when seemingly legitimate activity is fraudulent.

Visa recently said AI can serve only as one layer of defense against the new wave of scams and must be combined with transaction-data analysis, intelligence sharing across institutions and expert human investigations. As of July 20, 2026, the reports provided did not disclose the date of a specific case, the amount involved or a prevention rate. The key development is Visa's explicit position that fraud-prevention systems cannot rely solely on automated models.

Banks and merchants have long used passwords, one-time passcodes, selfies and voice biometrics to verify who is making a payment. Generative AI can now produce deepfakes, synthetic identities and voice clones at scale, weakening one-time login checks. Visa says that as AI agents begin shopping and paying on customers’ behalf, payment networks must determine both the scope of their authorization and whether their behavior is genuine, without compromising security or a smooth checkout experience.

PYMNTS published an interview with Visa executive James Mirfin on March 19, 2026. He said companies are continuously analyzing transaction patterns, device information and location data, while using tokenization and trusted-agent protocols to define authorization. The report did not disclose the value of fraud losses. Gartner estimates that by 2031, 50% of large financial institutions will consolidate responsibility for fraud prevention, identity and cybersecurity.

AI agents are moving beyond recommendations to make purchases, call APIs and buy computing resources, but existing checkout systems are still designed for humans. Stripe and payments blockchain Tempo therefore jointly developed the open Machine Payments Protocol, or MPP, allowing agents to receive payment requirements, authorize payments and access services in a single request. Visa’s integration with existing card networks is significant because it brings machine micropayments into mainstream payment and risk-control systems.

Stripe and Tempo released the MPP specification on March 18, 2026, alongside the launch of Tempo’s mainnet. Visa released a card-based specification and SDK the same day and enabled them on the Visa Acceptance Platform, with support for tokenization, authentication and agent identity verification. Visa said on July 14 that MPP had settled about $25,000 across roughly 115,000 transactions in the weeks since its mid-March launch, averaging less than 1 cent per transaction.

Although check usage continues to decline in the United States, check fraud remains a major area of financial crime. Visa said check fraud accounted for 30% of U.S. fraud losses in 2024. More significantly, the methods used in such crimes are crossing from paper to digital channels, threatening faster-payment systems and banks’ risk controls.

Visa recently warned that anomalies involving accounts, identities and transactions linked to check fraud have spread to faster digital payments. The company is deploying AI-powered image forensics and behavioral analysis to help banks identify altered checks and suspicious cross-channel transactions. The report did not disclose the actual amount lost or the date when the system became operational.