2026年4月28日上午9:00，WIRED报导FIDO联盟在Google与Mastercard协作下宣布成立两个工作小组，制定AI代理执行之支付与其他交易的产业标准，以避免未来AI代理替代人类下单时导致的灾难性风险。目标是建立跨产业可采用的保护底线，让授权机制更难被网钓、帐户劫持或坏人指令接管所破坏，并提高交易可验证性与授权可信度。
标准草案计划加入密码学工具，让数位服务能确认代理是否依「已验证的人」指令准确且合法执行；同时导入隐私保护架构，使使用者、商家与服务供应方能验证代理发起交易，同时最小化资讯外泄。FIDO CEO Andrew Shikiar指出，密码体系过去十余年在新威胁下已暴露出重大问题，现在AI代理互动同样站在类似的关键转折点，需要在更早期就建立基础原则避免重复走错路。
Google的Agent Payments Protocol（AP2）与Mastercard与Google共同开发的Verifiable Intent将为各方提供起步优势：AP2可密码学证明交易确为用户本人意图，Verifiable Intent让使用者在复杂支付生态中授权并控制代理行为。Stavan Parikh的示例是：指定AI在球鞋再上架且价格不高于100美元（USD 100）时自动下单。主导者Pablo Fourez、Parikh和其他代表都指出，AI代理发展速度很快，传统需要2至3年的标准化周期正被压缩，以降低对消费者与商户的高昂风险成本。
On Apr 28, 2026 at 9:00 AM, WIRED reported that the FIDO Alliance, with Google and Mastercard, announced two working groups to set standards for payments and other transactions executed by AI agents. The goal is to create an industry-wide baseline so AI-driven shopping and payment actions can avoid becoming a disaster from phishing, account takeover, or rogue instruction injection. The initiative is positioned to make authorization less forgeable and transactions more reliably verifiable.
The proposed standards include cryptographic mechanisms that let digital services check whether an agent is accurately and legitimately carrying out authenticated human intent, plus privacy-preserving designs so users, merchants, and service providers can validate agent-initiated transactions without excessive data exposure. FIDO CEO Andrew Shikiar said that password-era controls became inadequate over time, and AI agent commerce now faces a similar inflection point requiring foundational safeguards before mainstream adoption expands further.
Google’s Agent Payments Protocol (AP2) and Mastercard’s Verifiable Intent, co-developed with Google, are intended as practical starting points. AP2 proves cryptographically that a payment transaction was truly intended by the user, while Verifiable Intent gives users controlled authorization over agent actions within a multi-party payment ecosystem. Stavan Parikh gave the example of instructing an agent to buy sneakers only if they return in stock and price at USD 100 or less. With rapid AI adoption, the effort is treated as urgent because timelines that once took 2 to 3 years for standards work now need to be compressed to avoid high costs to consumers and merchants.