Industry partners propose open standard to help AI agents bypass anti-bot defenses

Meta, Walmart, Stripe, Sierra, Genesys, Rocket, NiCE, and Decagon are collaborating on an open standard to help AI agents interact with websites for commerce tasks such as booking flights, making dinner reservations, and ordering groceries. The protocol aims to distinguish legitimate agent activity from malicious bots, addressing a growing problem where websites block AI agents from completing user-requested tasks.

The issue has surfaced across multiple industries. Amazon recently blocked Meta’s Muse AI agent from browsing or making purchases on its retail site, while users have reported similar blocks on Walmart’s platform. Walmart, despite partnering with Meta, acknowledged that its website’s human-verification steps—such as CAPTCHA-like buttons—can disrupt agent workflows if interrupted, leading to failed transactions. Walmart stated it wants to be accessible via AI agents and attributed some blocks to unintentional misconfigurations.

Other companies have taken stricter stances. Yelp requires non-human traffic to access its platform through a paid data licensing program, effectively blocking agents without formal partnerships. eBay restricts unauthorized agents but does not prohibit all third-party shopping agents, focusing instead on preventing automated scraping and model training. Airlines like Delta and United have not explicitly blocked agents but cite security concerns and terms of service that prohibit automated activity without prior permission.

Cloudflare, a major provider of anti-bot and security services, may be unintentionally contributing to the problem. A change to its crawler defaults on September 15 shifted some sites to block AI agents on pages with ads, even if those sites previously allowed other types of bots. This adjustment was intended to let site owners block AI training while permitting other automated traffic, but it has had the side effect of disrupting agent workflows in some cases.