Mastercard has started testing in the United States a technology capable of indicating the probability that a transaction was initiated by an artificial intelligence agent. The tool is part of the expansion of Agent Pay and aims to give banks and merchants more context to authorize purchases made by autonomous systems.
The service assigns a probability score to transactions to help issuers differentiate operations made by AI agents from conventional e-commerce. In this first phase, the technology is available for testing in the American market.
Mastercard aims to combine signals related to agent behavior, merchant risk, transaction characteristics, credentials used and consumer profile. The goal is to identify legitimate automated activities and operations that may require additional verifications.
Banks gain more context about who is buying
The change becomes relevant because an agent can execute several purchases in sequence on behalf of the user. A system tasked with organizing a trip, for example, can book a ticket, a hotel and transportation at different establishments, behavior that traditional fraud prevention systems may consider unusual.
The technology is part of the Mastercard Agent Pay Trust Framework, a framework created to verify the agent's identity, the consumer's intent and the limits defined for a transaction.
The company also works with Cloudflare and Skyfire on technologies designed to identify agents and combine internet signals with information from the payments network.
The initiative accompanies the financial sector's preparation for the advance of so-called agentic commerce. Mastercard estimates that one in ten consumers may use AI agents regularly to make purchases by 2030.
