The end of checkout: Mastercard enables autonomous AI payments

The consolidation of the agentic economy

The definitive barrier between artificial intelligence as a consultant and artificial intelligence as an executor has fallen irreversibly. Mastercard's recent announcement, made on September 17, 2026, regarding a payment infrastructure designed specifically for autonomous agents, marks the beginning of the so-called agentic economy. This evolution shifts the paradigm of generative AI, which until now was limited to recommending or drafting, toward full transactional capability. In collaboration with the startup Alchemy, Mastercard allows AI agents to use virtual cards linked to pre-established parameters, eliminating the friction of manual checkout.

How does the trust architecture work?

The core of this innovation lies in so-called agentic tokens. Unlike traditional payment methods that rely on synchronous authentication—where every purchase requires an OTP code, biometrics, or human confirmation—this system operates through programmed delegation. The issuing bank encapsulates the user's intent into a digital token that acts as a smart contract: it contains spending limits, merchant whitelists, and category restrictions. The Mastercard network verifies this metadata in real-time before authorizing the charge. This architecture, described by the entity itself as a "card with identity, purpose, and trust," ensures that the agent cannot exceed the framework defined by the human.

From a technical perspective, this represents a fundamental shift in financial interoperability. Agents are no longer external entities requesting access to an account; they become issuers of validated instructions within the payment network itself. Sources close to Alchemy suggest that the "checkout button" is destined for obsolescence, as AI will be able to discover, compare, and acquire goods with a precision that exceeds human attention span.

The context: a race for invisible infrastructure

Mastercard's entry is a direct response to an ecosystem that has matured under the early leadership of Visa. Visa was a pioneer in partnering with Alchemy, setting a precedent that forced competitors to integrate similar capabilities to avoid losing relevance in the new value layer of the Internet. With the union of both giants, the vast majority of credit cards globally now have the technical capacity to enable autonomous agents.

Historically, this event is comparable to milestones such as the mass adoption of SSL protocols for e-commerce in the 90s or the transition to digital wallets (Apple Pay/Google Pay) a decade ago. However, there is a qualitative difference: in previous transactions, the user was always the engine of the action. Here, the user becomes a policy architect, delegating execution to an AI that acts on their behalf. We are witnessing the consolidation of an infrastructure that makes payment an "invisible function" of software.

Market implications and risks

The mass adoption of this technology carries challenges that the industry is only just beginning to articulate:

• Security and adversarial attacks: The attack surface expands. A malicious third party could attempt to manipulate the agent's context (prompt injection) so that it makes purchases outside authorized parameters or at fraudulent merchants. Security will no longer depend solely on encryption, but on the robustness of the language model guiding the agent.
• Legal vacuum and civil liability: Current jurisprudence still struggles to define the legal personality of autonomous agents. In the event of an erroneous transaction, legal liability between the user, the agent developer, and the payment network remains a gray area that courts will have to resolve in the coming years.
• Evolution of digital marketing: Brands will no longer compete solely for the human click. "Algorithmic preference" will become the new SEO. If an AI agent is programmed to buy "the product with the best value for money," brands will need to optimize their data to be readable and preferable by automated systems, not people.

In conclusion, we are heading toward a future where human interaction in e-commerce will be the exception, not the rule. AI has shifted from being a consultation tool to an agent with its own spending power, closing the loop on the full automation of personal life. The agentic economy is not only changing how we pay, but also the purchase decision process itself, forcing companies and regulators to rethink the foundations of digital trust.