FIDO Alliance, Google, and Mastercard Create Standard for AI Agents in Online Shopping

AI agents that make purchases for you are moving beyond the concept stage. The question is no longer "will they," but "how safe will they be" — and this is the question that FIDO Alliance, Google, and Mastercard are tackling.

Why This Became Urgent

Autonomous shopping is no longer an experiment. OpenAI launched Operator: an agent that controls a browser in real time, searches for products, selects items, and clicks "Pay" without human intervention. Google presented Agent Mode in Chrome, Anthropic is testing Computer Use, and dozens of startups are embedding similar features into search and browsers. The problem is that there is no single authentication standard for AI agents during transactions. Each company solves this independently — or ignores it altogether. While agent shopping remains a niche function, the risks are manageable. When this capability appears on every smartphone, the cost of lack of coordination will fall on users. Without a common protocol, obvious gaps emerge:

• An agent makes purchases without explicit user consent
• An attacker intercepts an agent's session and conducts unauthorized transactions
• An agent lands on a phishing store and transmits card details
• Multiple competing agents duplicate the same order
• The absence of spending limits leads to unforeseen expenses

Who Took on the Task

FIDO Alliance — a nonprofit organization that created passkey and passwordless authentication standards, which are now supported by Apple, Google, Microsoft, and most major banks — announced a partnership with Google and Mastercard. The goal: develop an open standard for authenticating AI agents when performing payment operations — a protocol that will link agent identification, user consent, and payment infrastructure into a single secure chain. This is not the first time FIDO Alliance has taken on a large-scale task.

A few years ago, SMS codes seemed an integral part of online banking. Today, passkeys are replacing them across all major platforms. It is precisely the partnership with Apple, Google, and Microsoft that turned passkeys from an academic project into real market change — without the participation of infrastructure players of this level, any standard remains a recommendation on paper.

Google controls Chrome and Android — most online purchases in the world pass through these platforms. Mastercard manages one of the largest payment networks, processing billions of transactions annually. Together, they close the entire chain: authentication → browser → transaction.

How Protection Might Work

The specific specifications are still in development, but the architectural logic is clear from previous FIDO Alliance standards: an open protocol available to any developer, not a proprietary solution from one corporation. Likely key elements:

• Delegated tokens — an agent receives a limited, single-use key rather than permanent access to card details
• Contextual verification — the payment system confirms that the operation is initiated by an authorized agent within pre-agreed parameters
• Explicit limits — the user sets a budget, permitted stores, and product categories
• Audit log — every action of the agent is recorded and remains available for review

If the standard is adopted by major platforms and payment networks, developers of agents will have to support it — this is how infrastructure standardization works: the rules are set by whoever controls the infrastructure.

What This Means

The race for autonomous AI agents is in full swing, and the industry is finally recognizing: the speed of deployment without safety standards creates real risks for users. The initiative by FIDO Alliance, Google, and Mastercard is an attempt to establish rules before agent transactions become massive and costly. If the standards take hold, having an agent order products or buy airline tickets will be safer than entering card details manually on an unfamiliar website.