Mastercard tests whether payment networks can tell humans from autonomous software

Mastercard is pushing deeper into the fast-growing world of “agentic” software, with a new effort aimed at answering a basic but tricky question for card payments: can a payment network reliably tell when a transaction is being initiated by a human versus by autonomous software?

In a report published by Yahoo Finance, the company described an “Agent Detection Score” that is part of its Agent Pay Trust Framework. The framework is positioned as an intelligence layer designed to support trust and risk evaluation as payments become increasingly automated, including scenarios where software agents take actions on a customer’s behalf.

At the center of Mastercard’s approach is the idea of scoring or assessing the nature of a transaction request, in order to inform what the network should treat as “normal” behavior. The challenge, as framed in the report, is that card networks have not previously faced the same kind of problem at scale: transactions that may be technically valid but initiated by non-human systems.

Mastercard’s description in the Yahoo Finance piece suggests the new score is meant to help close that gap by identifying indicates that may separate human-driven shopping from activity that originates from autonomous software. In practice, that could affect how payments are authorized and how risk controls are applied when software agents are involved.

The company’s focus on “trust” language is also notable. In payment systems, trust typically refers to the confidence level that the customer, the device or channel, and the transaction intent are legitimate. By extending trust evaluation toward automated agents, Mastercard is effectively treating software-driven payments as a new category of risk and verification, rather than simply another payment channel.

Mastercard’s move lands in a period when the payments industry is increasingly exploring how AI and software automation could change purchasing workflows. Software agents can aggregate data, make decisions, and execute actions quickly, which can be helpful to customers and businesses, but it can also complicate fraud detection and authorization decisions because traditional behavioral patterns may not match.

While the report indicates Mastercard is working on the Agent Detection Score within the Agent Pay Trust Framework, it does not, in the information available here, lay out the exact methodology behind the scoring, what data indicates it uses, or how it is intended to be rolled out across issuers and merchants.

What to watch next is whether Mastercard shares more specifics about the framework’s performance, the types of agent activity it can distinguish, and how the score will be integrated into existing authorization and risk systems. For the broader industry, the key question will be whether payments can add a reliable layer of agent verification without disrupting legitimate automated commerce.