FIDO Alliance Google and Mastercard Collaborate to Standardize Security and Authentication for Autonomous AI Agents in Commerce

The FIDO Alliance, a global industry consortium dedicated to reducing the world’s over-reliance on passwords, has announced the formation of two new working groups tasked with establishing security standards for "agentic AI." This initiative, bolstered by foundational contributions from Google and Mastercard, seeks to create a robust framework for validating and protecting transactions initiated by artificial intelligence agents on behalf of human users. As AI moves from passive assistants to active agents capable of making purchases and managing digital identities, the industry is racing to ensure these autonomous actions are secure, authenticated, and resistant to emerging cyber threats.

The move marks a significant shift in the digital security landscape. While traditional security models were designed for direct human-to-machine interaction, the rise of agentic AI introduces a paradigm where software agents act as intermediaries. Without standardized protocols, these agents could be vulnerable to hijacking, "prompt injection" attacks, or rogue instructions that lead to unauthorized financial transactions or data breaches. The FIDO Alliance aims to preempt these risks by developing a baseline of cryptographic protections that can be adopted across the global financial and technology ecosystems.

The Shift Toward Agentic Commerce and Its Inherent Risks

The concept of agentic AI refers to systems that do not merely provide information but take meaningful actions to achieve a goal. In a commercial context, this might involve an AI agent monitoring price fluctuations for a specific product and executing a purchase once certain conditions are met. While this offers unprecedented convenience, it also creates a new attack surface for cybercriminals. Existing security measures, such as multi-factor authentication (MFA) designed for human confirmation, are often ill-suited for background processes carried out by autonomous code.

According to Andrew Shikiar, CEO of the FIDO Alliance, the current security infrastructure is at a critical juncture. He notes that the industry is currently at a "precipice" similar to the one faced decades ago when the internet began to rely on passwords—a system that eventually proved to be the primary vector for data breaches. By establishing foundational principles for agentic interactions now, the alliance hopes to avoid the legacy security flaws that have plagued the connected economy for the last thirty years.

The risks associated with unstandardized AI agents are multifaceted. Malware could potentially intercept agent instructions, or online impersonation could lead a service provider to believe a rogue agent is acting with a user’s legitimate authority. Furthermore, "account takeovers" could evolve into "agent takeovers," where a bad actor gains control of a user’s digital representative to drain bank accounts or access sensitive personal information.

Technical Frameworks: Google’s AP2 and Mastercard’s Verifiable Intent

To jumpstart the standardization process, Google and Mastercard have contributed significant intellectual property and open-source tools to the FIDO Alliance. These contributions form the technical backbone of the proposed standards, focusing on two key areas: cryptographic verification and user authorization.

Google has introduced the Agent Payments Protocol, or AP2. This protocol is designed to provide a mechanism for digital services to cryptographically verify that a user intended for a specific transaction to occur, even if that transaction was initiated by an AI agent in the background. AP2 uses public-key cryptography to sign transactions, ensuring that the instructions have not been tampered with and that they originate from a trusted source.

Complementing this is Mastercard’s Verifiable Intent framework, which was co-developed with Google to integrate seamlessly with AP2. Verifiable Intent focuses on the "intent" aspect of the transaction. It allows users to set granular permissions and boundaries for their agents. For example, a user could authorize an agent to spend up to $100 on a specific category of goods without further intervention, but require manual approval for anything exceeding that limit. This framework ensures that the agent remains under the user’s ultimate control, providing a secure mechanism for authorization.

Stavan Parikh, Google’s Vice President and General Manager of Payments, emphasizes that the goal is to provide "cryptographic proof" of authorization while maintaining user privacy. The frameworks utilize "selective disclosure," a privacy-preserving technique where different entities in the payment ecosystem—such as merchants, banks, and platform providers—only see the specific data points necessary to fulfill their role in the transaction. This prevents the over-sharing of sensitive user data while still ensuring the transaction is legitimate.

A Chronology of Authentication and the Path to AI Standardization

The development of these AI standards is the latest chapter in a long history of authentication evolution. Understanding this timeline is essential to grasping the importance of the FIDO Alliance’s current mission:

• 1960s – 2000s: The Password Era: The early digital age relied almost exclusively on alphanumeric passwords. As the internet grew, the limitations of passwords—such as being easily guessed, phished, or stolen in bulk—became the primary security challenge for the global economy.
• 2012: The Birth of FIDO: The FIDO (Fast IDentity Online) Alliance was formed to address the lack of interoperability among strong authentication technologies. Founding members sought to move beyond passwords toward hardware-backed security.
• 2014 – 2019: FIDO1 and FIDO2: The alliance released standards for Universal Second Factor (U2F) and later FIDO2, which enabled passwordless logins using biometrics (like fingerprints or facial recognition) and security keys.
• 2022: The Rise of Passkeys: Apple, Google, and Microsoft announced expanded support for FIDO standards through "passkeys," allowing users to sync their cryptographic credentials across devices, effectively making passwordless authentication mainstream.
• 2023 – 2024: The Generative AI Explosion: The rapid adoption of Large Language Models (LLMs) and agentic frameworks created an immediate need for delegated authority. Security experts realized that passkeys, while secure for humans, did not yet account for autonomous agents acting on a human’s behalf.
• 2025: FIDO Agentic AI Working Groups: The launch of the new working groups marks the official start of industry-wide standardization for the next generation of the "agentic economy."

Supporting Data and the Economic Necessity of AI Security

The urgency behind these new standards is driven by the projected growth of the AI market and the persistent threat of financial fraud. Industry analysts suggest that the market for AI agents could reach hundreds of billions of dollars by the end of the decade as businesses integrate autonomous systems into supply chains, customer service, and personal finance.

Data from recent cybersecurity reports highlights the stakes:

• Digital Payment Fraud: Global losses from payment fraud are expected to exceed $40 billion annually by 2027. Without secure agent protocols, this figure could spike as attackers target autonomous transaction streams.
• Phishing Resilience: FIDO-based authentication has been shown to reduce successful phishing attacks to near zero. Applying these same cryptographic principles to AI agents is seen as the only viable way to protect the future of automated commerce.
• AI Adoption Rates: A significant percentage of enterprise leaders cite "security and privacy concerns" as the primary barrier to adopting agentic AI. Standardizing these protections is viewed as a prerequisite for widespread commercial confidence.

By creating a "protective baseline," the FIDO Alliance aims to lower the barrier to entry for smaller companies and developers who may not have the resources to build complex security infrastructures from scratch. Interoperability ensures that an agent developed by one company can securely interact with a payment gateway managed by another, creating a frictionless and secure ecosystem.

Industry Reactions and Broader Implications

The announcement has garnered support from various sectors of the technology and financial industries. While Google and Mastercard are leading the initial contribution, the standards are expected to involve input from hundreds of FIDO member companies, including banks, hardware manufacturers, and software developers.

Industry analysts suggest that the success of this initiative will depend on how quickly these standards can be finalized and implemented. Unlike the transition to passkeys, which took several years to gain momentum, the "AI summer" is moving at a pace that demands rapid iteration. The inclusion of open-source tools from Google and Mastercard is a strategic move to accelerate this timeline.

The broader implications of these standards extend beyond simple retail transactions. If successful, the FIDO framework for agentic AI could be applied to:

1. Healthcare: AI agents managing patient data or scheduling medical appointments while ensuring HIPAA compliance through cryptographic verification.
2. Corporate Procurement: Autonomous agents managing B2B supply chains, placing orders, and verifying invoices within pre-approved budgetary and security parameters.
3. Smart Cities: Agents managing energy consumption or autonomous vehicle payments for tolls and charging stations without requiring constant human intervention.

The "selective disclosure" aspect of the protocol is particularly noteworthy for privacy advocates. In an era of increasing data regulation, such as the GDPR in Europe and the CCPA in California, the ability to validate a transaction without exposing the user’s entire digital identity is a significant step forward for "privacy by design."

Conclusion: Securing the Future of Autonomous Interactions

As AI agents become a mainstream feature of the digital experience, the work of the FIDO Alliance, Google, and Mastercard represents a proactive effort to build security into the foundation of the technology rather than bolting it on as an afterthought. By moving away from human-centric authentication models toward delegated cryptographic authority, the industry is preparing for a world where "agentic interaction" is the norm.

The goal, as stated by Google’s Stavan Parikh, is to ensure that when a consumer instructs an AI agent to perform a task—whether it is finding a sold-out pair of sneakers or managing a complex financial portfolio—the outcome is exactly what the user intended. Through transparency, accountability, and robust cryptographic proof, the new FIDO standards aim to foster the trust necessary for the next phase of the global digital economy. The working groups are expected to begin their sessions immediately, with initial draft standards likely to be circulated among stakeholders within the coming year.