AI代理面临网站准入障碍 行业联手制定通信标准

随着Meta Muse、ChatGPT等个人AI代理的推出，这些工具能够代用户完成订机票、预订餐厅等在线任务，但正面临来自主要电商平台的访问限制。1Amazon已明确阻止Meta Muse AI代理在其零售网站上浏览或购物，1Walmart在与Muse的伙伴关系中也遇到问题，用户在验证人类身份的步骤中被拦截。1

网站对AI代理的态度各不相同。Yelp明确表示，除非代理通过其数据许可计划付费，否则不允许非人类流量。1Delta则表示如果开放AI代理访问，需要"从安全和客户体验角度出发"。1部分网站的屏蔽出于有意为之，但也有传统反爬虫措施无意中误伤合法代理的情况。1

为解决这一问题，Meta与Walmart、Stripe、Sierra、Genesys、Rocket、NiCE和Decagon等行业伙伴正在开发针对AI代理与企业通信的开放标准，以期规范两者的交互方式，区分合法和恶意机器人。1此外，Cloudflare在9月15日改变了爬虫默认设置，允许网站所有者在屏蔽AI训练的同时允许其他类型的机器人，为行业提供了更细粒度的控制工具。1

Personal AI agents like Meta Muse and ChatGPT Dots are designed to complete tasks on behalf of users—such as booking flights or reserving restaurant tables—but they are encountering significant barriers as major websites move to block them 1. Amazon has explicitly prevented Muse from browsing or making purchases on its retail platform, while Walmart, eBay, and other e-commerce sites have reported similar blocking issues 1. Some of these barriers are intentional, while others result from traditional anti-scraping measures that inadvertently catch AI agents 1.

Walmart has faced particular friction in its partnership with Muse, with users encountering human identity verification steps that prevent the agent from proceeding 1. Yelp has taken an explicit stance, stating that it will not permit non-human traffic unless the agent operator pays through its data licensing program 1. Delta has indicated that granting AI agents access to its systems would require careful consideration "from a security and customer experience perspective" 1. These responses highlight the tension between enabling useful AI-powered services and protecting business interests and user security.

To address these challenges, Meta is collaborating with industry partners including Walmart, Stripe, Sierra, Genesys, Rocket, NiCE, and Decagon to develop an open standard for how AI agents communicate with businesses 1. This standardization effort aims to distinguish between legitimate agents and malicious bots, creating a framework that could allow websites to grant controlled access to trusted AI systems while maintaining security. Meanwhile, Cloudflare modified its default crawler settings on September 15, enabling website owners to block AI training traffic while permitting other types of bots 1.