Technical answer: The FIDO Alliance is developing new industry standards to secure AI agents, focusing on cryptographic verification of user intent to prevent agent hijacking and unauthorized transactions.
FIDO Alliance Agentic AI Standards: Securing Transactions and User Intent
Category: Cybersecurity / AI Policy | Technology: Agentic AI, FIDO Authentication, Cryptographic Protocols (AP2) | Article Type: News Analysis / Technical Explanation
Quick Context: The FIDO Alliance, with contributions from Google and Mastercard, is launching working groups to standardize the authentication and protection of transactions initiated by AI agents on behalf of human users.
Table of Contents
Technical Evaluation Path
Identify Agentic Use Cases: Determine where AI agents are currently (or will be) performing autonomous actions such as payments, booking, or data exchange.
Review Exposure to Hijacking: Assess the risk of "rogue instructions" where an agent could be manipulated into unauthorized transactions.
Analyze Protocol Contributions: Evaluate Google’s AP2 and Mastercard’s Verifiable Intent framework for compatibility with existing identity stacks.
Determine Cryptographic Requirements: Establish the infrastructure needed to support selective disclosure and cryptographic proof of intent.
Monitor Working Group Progress: Track the FIDO Alliance's development of these standards to ensure future interoperability across merchants and providers.
Key Facts From the Source
Standardization Body: The FIDO Alliance is forming two new working groups dedicated to "Agentic AI."
Key Contributors: Initial technical contributions are provided by Google and Mastercard.
Core Objective: To create a protective baseline for validating and protecting transactions carried out by AI agents to prevent phishing and hijacking.
Protocols Introduced: * Agent Payments Protocol (AP2): A Google-led mechanism for cryptographically verifying user intent.
Verifiable Intent Framework: A Mastercard-led (co-developed with Google) mechanism for user authorization and control.
Privacy Focus: The standards aim for "selective disclosure," ensuring participants in the ecosystem only see data relevant to their specific role in the transaction.
Technical Context or Problem Background
The "Agentic AI" paradigm shifts activity from humans performing tasks to agents acting on a human's behalf. Traditional security models, such as standard passwords or basic MFA, were not designed for autonomous agents that may need to trigger actions (like a purchase) days or weeks after the initial human instruction.
Without a standardized cryptographic link between the user’s original intent and the agent’s eventual action, the ecosystem is vulnerable to agent hijacking. This occurs when a bad actor intercepts or modifies the instructions given to an AI, leading to unauthorized financial or data transactions.
First Checks Before Production Use
Before adopting agentic AI payment or transaction workflows based on these emerging standards, organizations should verify the following:
Intent Integrity: How is the original "user intent" captured and signed?
Phishing Resistance: Does the agent authorization mechanism rely on legacy credentials, or is it tied to FIDO-compliant hardware/biometrics?
Interoperability: Will the merchant or payment processor support the selective disclosure proofs generated by the agent?
Revocation Capability: Can a user transparently view and revoke a "long-running" instruction given to an agent before the transaction occurs?
Practical Takeaways
Move Toward Cryptographic Proof: Shift away from simple API keys or "allow-all" permissions for AI agents toward signed, intent-based authorizations.
Prepare for "Selective Disclosure": Implementation will likely require systems that can process zero-knowledge or minimal-data proofs to maintain user privacy.
Standardization Speed: Because AI adoption is outpacing traditional standards timelines (2–3 years), expect rapid iterations of these protocols in the open-source community.
Accountability Models: Organizations must prepare for dispute resolution frameworks specifically designed for "agent-initiated" errors or rogue behaviors.
Comparison of Contributed Protocols
Protocol / Framework
Primary Function
Advantage
Current Source/Status
Agent Payments Protocol (AP2)
Cryptographic verification of user intent
Ensures the transaction matches the user's specific original instruction.
Google (Open Source Contribution)
Verifiable Intent
Authorization and control mechanism
Provides a secure "handshake" for users to delegate power to agents.
Mastercard / Google (Open Source)
Verification Checklist
Are the AI agents in your environment performing actions that involve financial or sensitive data exchange?
Yes
No
Needs Review
Is there a cryptographic link between the user's initial prompt and the final transaction execution?
Yes
No
Needs Review
Have you reviewed the AP2 or Verifiable Intent open-source documentation for potential integration?
Yes
No
Needs Review
Does your current security architecture account for "long-running" agent authorizations?
Yes
No
Needs Review
Related Articles or Next Steps
Related Linked Tags
FIDO Alliance
Agentic AI
Google AP2
Mastercard Verifiable Intent
Cybersecurity Standards
AI Authentication
Payment Security
Phishing Resistance
Cryptographic Protocols
Digital Identity
Images or Diagrams on Demand
Visual 1
Purpose: Illustrate the "Agent Hijacking" threat vector vs. a secured FIDO/AP2 transaction.
Possible file name: agent-hijacking-vs-secure-intent-verification.webp
Possible caption: Comparison of unsecured agent workflows and the proposed FIDO Alliance cryptographic verification layer.
When to request it: Useful for high-level briefings on the necessity of these new standards.
Visual 2
Purpose: Map the "Selective Disclosure" flow between User, Agent, and Merchant.
Possible file name: selective-disclosure-payment-flow.webp
Possible caption: How different players in the ecosystem only see data relevant to their specific role.
When to request it: Use this when explaining the privacy-preserving aspects of the protocol to compliance or privacy teams.
60-Second Reading Overview
The FIDO Alliance has launched two working groups to establish industry standards for Agentic AI, focusing on how AI agents perform transactions on behalf of humans. As AI agents move into mainstream use—performing tasks like autonomous purchasing—they introduce risks such as agent hijacking and rogue behavior. To combat this, Google and Mastercard have contributed open-source protocols: AP2 for cryptographic verification of intent and the Verifiable Intent framework for user authorization. These standards aim to create a "phish-proof" baseline that ensures transparency, privacy through selective disclosure, and clear accountability for agent-initiated actions. For technical teams, this represents a shift toward intent-based cryptographic security in the AI-driven connected economy.
Search & Classification Metadata
Category: Cybersecurity / AI Infrastructure
Technology: FIDO2, AP2, Verifiable Intent, AI Agents, Cryptography
Object Type: Technical Standard / Security Advisory
Use Case: Securing autonomous AI agent transactions and preventing unauthorized agent behavior.
Article Type: News Analysis / Technical Explanation
Applies To: Financial services, e-commerce, AI developers, identity providers, and cybersecurity architects.
Search Intent: FIDO agentic AI standards, Google AP2 protocol explained, Mastercard Verifiable Intent framework, securing AI agent payments, AI agent hijacking prevention.
Related Documentation
FIDO Alliance Official Announcement: [Agentic AI Working Groups]
Google Open Source: [Agent Payments Protocol (AP2) Documentation]
Mastercard Technical Framework: [Verifiable Intent Documentation]
FAQ
What is the main technical takeaway of this FIDO Alliance initiative?
The initiative aims to standardize how we prove a human actually authorized a specific action performed by an AI agent, using cryptography instead of just basic API permissions.
Who is affected by these new agentic AI standards?
Anyone building, deploying, or accepting payments from AI agents, including developers, merchants, and payment processors.
What are AP2 and Verifiable Intent?
AP2 (Agent Payments Protocol) is a Google contribution for verifying intent; Verifiable Intent is a Mastercard/Google framework for users to authorize and control what their agents can do.
How does this prevent "agent hijacking"?
By requiring cryptographic proof that matches the user's original intent, the protocol makes it significantly harder for a bad actor to inject "rogue instructions" into an agent's workflow.
Will these standards protect user privacy during payments?
Yes, the frameworks use "selective disclosure," meaning a merchant might see that a payment is authorized without seeing the full context of the user's private instructions to the agent.
Is this technology ready for immediate implementation?
While the open-source tools (AP2 and Verifiable Intent) are available, the FIDO Alliance working groups are currently refining them into a formal industry standard, which requires further testing and use-case validation.
Contact / Support
Need professional technical support, implementation guidance, or troubleshooting help?
Contact your internal technical team, vendor support, or a qualified specialist before applying changes to production systems.