Mastercard: 2030 year of AI agentica, when software will pay for us

"The goal is to anticipate fraud," Corti continues, "We can give a scoring to the transaction. Probably the human transaction will behave differently from that of a genuine agent, and therefore the intelligent infrastructure that sees a transaction that is out of the usual pattern will raise the score that the bank will then decide to use together with other key data.

Mastercard is collaborating with leading artificial intelligence and commerce leaders, including Stripe, Google and Ant International's Antom, to make secure agent transactions accessible and scalable. By the holiday season, all US Mastercard cardholders will be enabled in the Mastercard Agent Pay programme, with a global rollout to follow. Citi and U.S. Bank Mastercard cardholders will be able to experience AI-powered shopping, with vendors and agent commerce enablers such as PayOS and other industry players coming on board.

Together with the FIDO Alliance and its Payments Working Group, Mastercard and other industry leaders are developing a verifiable credential standard for payments that confirms transaction details such as amount, merchant and product. This ensures that all parties involved in a transaction can be certain that it has been approved by the acquirer, paving the way for a safer, smoother and more reliable basis for payments agents.

Among the tools is Agent Toolkit, available from Mastercard Developers, a toolkit that enables AI assistants and agent tools to access and interpret Mastercard API documentation using structured, machine-readable content. This supports integration with platforms such as Claude, Cursor and GitHub Copilot.

On the attack front, Corti explains, the strategy is based on acquisitions: 'In the last five years we have made acquisitions worth about 17.5 billion. For example Recorded Future, for $2.65 billion, which is able to monitor the dark web. We receive 200 attacks per minute, so our strategy is to prevent, to play the attack in a logic of open innovation, that is, looking for the most innovative companies on the market'.

According to the report, cybercrime could cost the world over $15bn by 2029, and where the average cost of a data breach is already $4.88m, adaptive AI and threat intelligence, as well as quantum-resistant technologies, will define the next generation of digital security.

On the Atm cash withdrawal front, Mastercard has launched a project with NCR Atleos and ITCARD that allows smartphone withdrawals via CDCVM, without card or PIN: the user authenticates on the phone and initiates the contactless transaction, 20% faster. The trial is underway in Poland and global roll-out is planned for 2026.