Amazon Blocks Meta's Muse Agent From Shopping on Its Site, Citing Unauthorized Access

Amazon has blocked Meta's Muse AI agent from buying goods on its behalf, escalating a fight over who controls the checkout page when software, not a person, does the shopping. A popup began appearing Sunday for Muse users, telling them that "continued access by an unauthorized AI agent violates Amazon's Conditions of Use, to which our customers have agreed." The message surfaced first in reporting by GeekWire, which also said Meta did not give Amazon advance notice before Muse started reaching into the retailer's storefront.

The mechanics matter as much as the message. Muse is built to act for a user: it browses listings, picks items and completes a purchase without the shopper tapping through each page. Amazon's Conditions of Use are written for human customers and the automated systems it authorizes, such as its own price-tracking and advertising tools. An outside agent completing transactions on someone else's account sits in a gap those terms never explicitly filled, and Amazon is now treating it as a violation rather than a feature.

The timing is awkward for Meta, which has spent the past year pushing agents from chat and research tasks into transactions. Muse was pitched as a way to hand off errands, and shopping is the most obvious errand there is. Losing Amazon removes the single largest U.S. retail catalog from the agent's reach, at least for now, and it hands rivals a talking point: an assistant that cannot buy what you ask it to buy is a demo, not a product.

There is also a competitive subtext. Amazon runs its own AI shopping features, including Rufus, and has every incentive to keep third-party agents from inserting themselves between its customers and its listings. Retailers have watched for years as platforms tried to intermediate their traffic; an agent that picks products can steer demand, capture data and, in theory, charge for placement. Amazon's popup is a line drawn before that becomes normal.

Meta has not said publicly how it will respond, and it is unclear whether the block is technical, contractual or both. The company could negotiate an approved-agent program with Amazon, route purchases through a different interface, or simply tell users to finish checkout themselves. Amazon, for its part, has not outlined what an authorized agent would look like or what it would cost to become one.

That gap is the story to watch. If Amazon will not let outside agents transact, every assistant maker faces the same wall, and the industry needs a shared answer on identity, liability and payment when software spends a human's money. The next move is Meta's, but the precedent belongs to Amazon.