The global technical body EMVCo has officially released a draft framework designed to establish a secure, interoperable, and scalable foundation for card-based agentic payments, marking a significant milestone in the evolution of autonomous commerce. As artificial intelligence (AI) increasingly transitions from a tool for information retrieval to an active participant in economic exchange, the need for standardized protocols has become paramount. Agentic payments—transactions initiated by AI agents on behalf of human consumers—present unique challenges to traditional payment architectures, particularly regarding the verification of intent, the management of liability, and the maintenance of transaction state over time.
EMVCo, which is owned by American Express, Discover, JCB, Mastercard, UnionPay, and Visa, serves as the primary standard-setting organization for the global payments industry. Its move to address agentic commerce reflects a proactive effort to ensure that as machines begin to make purchasing decisions, the underlying financial infrastructure remains robust against fraud while providing a seamless experience for merchants and issuers. The proposed framework introduces a novel concept known as Intent Services, a shared coordination layer that allows all participants in a transaction to track and update the authorization status of an AI agent in real-time.
Understanding the Shift to Agentic Commerce
To appreciate the significance of EMVCo’s new framework, one must understand the shift from traditional e-commerce to agentic commerce. In a standard digital transaction, a human user navigates a website, selects an item, and manually provides authorization through a password, biometric scan, or a one-time passcode. In the agentic model, a consumer delegates this authority to an AI agent. For example, a homeowner might authorize an AI-integrated smart home system to monitor grocery levels and purchase milk whenever the supply is low, within a specific budget.
While this offers unprecedented convenience, it introduces a "delegation of intent" problem. Unlike a human-initiated transaction, where intent is expressed at the moment of sale, agentic intent is often persistent, cumulative, and managed over a duration. This creates a visibility gap for financial institutions: how can an issuer verify that a specific transaction was truly authorized by the cardholder weeks after the initial permission was granted? EMVCo’s research into this sector highlighted that existing cryptographic solutions, while useful for one-time proofs, are insufficient for managing complex, long-term states such as recurring budgets or post-transaction activities like returns and disputes.
The Core Innovation: Intent Services and State Management
The centerpiece of the draft framework is the introduction of Intent Services. This is envisioned as a shared layer that enables payment participants—including the consumer’s bank (the issuer), the merchant, and the AI agent itself—to register, retrieve, and update consumer-authorized intent.
Under this framework, when a consumer authorizes an agent to act on their behalf, that intent is registered within the Intent Services layer. This registration includes specific parameters such as the duration of the authorization, the maximum spending limit, and the types of goods or services permitted. During a transaction, the merchant or the issuer can query this shared state to confirm that the agent is acting within the bounds of its delegated authority.
Crucially, this framework is designed to be complementary to existing security measures. It does not replace cryptographic protocols like Verifiable Intent, which use digital signatures to prove a consumer’s initial consent. Instead, Intent Services provides the coordination point necessary to interpret how that intent evolves. For instance, if an agent has a $500 monthly budget and has already spent $450, the Intent Services layer provides a consistent "state" that informs the issuer that only $50 remains available for the next transaction, regardless of which merchant the agent visits.
Chronology of Development and Industry Collaboration
The release of this draft framework is the result of intensive study by EMVCo’s dedicated Agentic Payments Task Force, which was established to investigate the impact of autonomous software on payment ecosystems. The timeline of this initiative reflects a rapid response to the generative AI boom that began in late 2022.
- Late 2023: EMVCo identifies agentic commerce as a priority area for standardization, noting the lack of a unified way to communicate "machine-to-machine" payment intent.
- Early 2024: The Agentic Payments Task Force begins consultations with major stakeholders, including EMVCo Associates and Subscribers, to identify the primary friction points in autonomous spending.
- Mid-2024: Collaborative efforts are formalized with other global standards bodies, including the FIDO Alliance (focused on authentication), the OpenID Foundation, the OpenWallet Foundation, and the World Wide Web Consortium (W3C).
- September 2024: EMVCo releases the draft "Framework for Secure, Interoperable and Scalable Card-Based Agentic Payments" for public comment.
- September 30, 2024: The deadline for industry feedback, after which EMVCo will refine the framework for potential integration into technical specifications.
This chronology demonstrates the industry’s recognition that the technology is moving faster than the regulatory and technical standards currently in place. By involving organizations like the W3C and FIDO Alliance, EMVCo is ensuring that the payment layer of AI agents is compatible with the broader web and identity authentication standards used across the globe.
Supporting Data: The Economic Imperative for Standardization
The push for a standardized agentic payment framework is supported by significant economic projections. According to industry analysis from firms like Gartner and Juniper Research, the "machine customer" market is expected to be one of the most significant growth drivers for digital commerce over the next decade. Some estimates suggest that by 2028, autonomous AI agents could influence or directly execute up to 20% of all digital commerce transactions.
Furthermore, data from payment processors indicates that "card-not-present" (CNP) fraud remains a multi-billion dollar challenge. Without a standardized framework like EMVCo’s, agentic payments could exacerbate this issue. If every AI agent uses a proprietary method for proving authorization, issuers will find it nearly impossible to distinguish between a legitimate autonomous purchase and a sophisticated bot-driven fraud attack. Standardizing the "Intent" data fields allows for better risk scoring and reduces the likelihood of false transaction declines, which currently cost merchants billions in lost revenue annually.
Official Responses and Strategic Vision
The leadership at EMVCo has emphasized that this framework is about building a foundation of trust. Junya Tanaka, the EMVCo Executive Committee Chair, stated that card-based agentic payments require a "globally interoperable foundation" that all parties—consumers, merchants, and issuers—can rely upon. Tanaka noted that the publication of this draft is an "important step forward" in establishing a consistent approach for communicating intent alongside payment data.
Industry analysts have reacted positively to the draft, noting that EMVCo is wise to look beyond the transaction itself. By focusing on the "lifecycle and state information" of intent, the organization is addressing the long-term management of AI permissions, which is where most legal and financial disputes are expected to arise.
The framework also outlines future capabilities that go beyond simple transaction processing. These include:
- Know Your Agent (KYA): Similar to Know Your Customer (KYC) regulations, KYA would allow payment participants to identify the specific AI agent involved in a transaction and verify its attributes and reputation.
- Agentic Transaction Indicators: These are specialized signals that would inform an issuer’s fraud detection system that a transaction is being performed by an agent rather than a human, allowing for different risk parameters to be applied.
Broader Impact and Implications for the Future of Finance
The implications of EMVCo’s proposed framework extend far beyond the technical specifications of a credit card transaction. If successfully implemented, this framework will provide the "rules of the road" for the autonomous economy.
For merchants, the framework offers a path to lower friction. If an AI agent can prove its intent and budget through a standardized EMVCo service, merchants can accept these payments with higher confidence, reducing the need for intrusive step-up authentication that might break the autonomous flow. This is particularly relevant for "invisible payments" in smart cities, automated logistics, and subscription-based IoT devices.
For consumers, the framework provides a much-needed layer of control. By standardizing how intent is registered and retrieved, consumers can theoretically manage all their AI agent permissions in one place. They could see, in real-time, which agents have access to their funds and revoke or modify those permissions instantly across different banks and merchants.
For the broader payment ecosystem, EMVCo anticipates that this framework will inform updates to other critical technologies. These include EMV 3-D Secure (3DS), which provides an extra layer of security for online purchases; EMV Payment Tokenization, which replaces sensitive card data with unique identifiers; and EMV Secure Remote Commerce (SRC), which streamlines the checkout process. By integrating agentic logic into these existing pillars, EMVCo is ensuring that the transition to AI-driven commerce is an evolution of the current system rather than a disruptive and risky overhaul.
As the September 30 deadline for feedback approaches, the global financial community is closely watching the development of these standards. The final framework will likely serve as the blueprint for how billions of dollars in autonomous transactions are secured in the coming years, bridging the gap between human financial intent and machine execution.
