Artificial intelligence agents are poised to become personal shoppers for millions of people, automatically hunting down products and completing purchases on your behalf. But before this becomes routine, major technology companies and security organizations are scrambling to build safeguards that prevent these digital assistants from being hijacked by criminals or malfunctioning in ways that drain bank accounts and damage consumer trust.
New Alliance Tackles AI Payment Security
The FIDO Alliance, an authentication-focused industry association, announced this week that it is launching working groups dedicated to developing security standards for AI agent transactions. The initiative includes contributions from Google and Mastercard, two companies that recognize the urgent need for protective frameworks as agentic AI moves from experimental technology into mainstream commercial use.
The fundamental challenge is straightforward but consequential. Existing security models were designed for transactions initiated directly by humans, not for transactions carried out by artificial agents acting on a person’s behalf. Andrew Shikiar, CEO of the FIDO Alliance, told WIRED that this represents a critical moment similar to the early days of internet commerce. Just as security foundations for the connected economy proved inadequate decades ago, the current infrastructure was never built to contemplate the risks of autonomous agent activity.
How the New Security Standards Would Work
The standards being developed aim to create multiple layers of protection. First, they would establish mechanisms that cannot be easily phished or compromised by attackers seeking to redirect agent instructions toward unauthorized purchases. Second, they would include cryptographic tools allowing digital services to verify that an agent is accurately following the authentic instructions of its human owner. Third, they would incorporate privacy-preserving frameworks that let users, merchants, and service providers validate transactions while maintaining appropriate confidentiality.
Google is contributing what it calls the Agent Payments Protocol, or AP2, which cryptographically verifies that a user genuinely intended for a given agent-initiated transaction to occur. Mastercard is providing the Verifiable Intent framework, developed in collaboration with Google, which serves as a secure mechanism for users to authorize and control agent actions. Stavan Parikh, Google’s vice president and general manager of payments, explained that the goal is to provide cryptographic proof of authorization while keeping sensitive information private through selective disclosure. Different players in the payment ecosystem would see only the information relevant to their specific role, ensuring that transactions proceed efficiently while maintaining security.
The companies illustrated this concept with a practical example. Imagine a shopper who finds sneakers sold out but wants an AI agent to purchase them if they return to stock at $100 or less. The new security standards would authenticate this instruction and ensure transparency so that when the shoes finally become available, the consumer receives exactly what they authorized at the price they specified.
Racing Against the Speed of AI Adoption
Developing technical standards that work across industries and facilitate interoperability typically requires years of careful work. Industry representatives emphasized, however, that the rapid advancement of agentic AI cannot wait for a traditional standards timeline. Pablo Fourez, Mastercard’s chief digital officer, noted that the technology is evolving extremely fast, compressing standards development timelines that in the past might have taken two or three years.
Establishing baseline protections is essential both for promoting consumer trust and enabling broader adoption of AI-powered tools. Whether people actively choose to use AI agents or not, their proliferation means that minimum safeguards become necessary as a matter of practical necessity. If systems fail to include fundamental security measures, the cost to cardholders and merchants from fraud and unauthorized transactions could be substantial. Mastercard stated it will always have the cardholder’s back but emphasized that preventing bad actors from exploiting vulnerabilities in the first place is far preferable to dealing with fraud after the fact.
The next phase of work will require building out practical examples and use cases to ensure the technology functions reliably in real world conditions. Then comes the harder challenge of achieving widespread adoption and support of these protocols at scale across users, platforms, merchants, payment providers, and other sectors. The race is on to establish these foundational principles before agentic commerce becomes too entrenched in systems that lack proper security safeguards.



