Mastercard built trust infrastructure for AI commerce

Mastercard launched Agent Pay earlier this year through OpenAI's Instant Checkout, making it the first payment system designed specifically for AI agents that transact on your behalf. The technical architecture matters here: "agentic tokens" create a digital credential that links each transaction to specific agent authorization. So when an AI books your flight or orders groceries, there's a verifiable trail showing which agent acted, what permissions it had, and where the money went.

This connects directly to what I've been writing about—the governance vacuum when AI moves from tool to actor. You've got systems that can initiate transactions at scale, and traditional payment rails weren't built for that. Mastercard's approach embeds authorization into the transaction layer itself, which means you're not retrofitting accountability after the fact.

For legal and product teams, this is the kind of infrastructure decision that shapes your agent strategy. When an AI system can spend money autonomously, "who authorized this" becomes a compliance question, a fraud question, and a user trust question all at once. Building that audit trail into the payment mechanism—rather than bolting it on later—is how you make autonomous commerce actually work.

Citi and US Bank cardholders can use it now; all U.S. Mastercard holders by end of the year.

https://www.mastercard.com/global/en/news-and-trends/stories/2025/agentic-commerce-momentum.html