Amazon blocks Meta's Muse AI agent from shopping on its site

Amazon has blocked Meta’s new Muse personal AI agent from shopping on Amazon.com, after Meta declined a request to keep the agent away from the store [1]. Since Sunday night, anyone who asks Muse to buy something on Amazon has been met with a wall: a notice that continued access by an unauthorized AI agent violates Amazon’s Conditions of Use [1][2]. It is the clearest sign yet that the fight over agentic shopping will be settled not by what an agent can do, but by whether the sites it visits agree to let it in.

What was announced

Amazon says it cut off Muse after trying, and failing, to get Meta to voluntarily exclude Amazon.com from the agent’s reach [1]. According to Amazon, Meta never told it that Muse would access the store, the agent does not identify itself as it browses, and it appears to capture and store customer credentials — a combination the company says creates privacy and security risks for its shoppers [1]. As of Sunday night, users who tried to shop Amazon through Muse were shown a pop-up reading, “Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed” [1].

An Amazon spokesperson framed the move as a matter of consent rather than capability: “We think it’s fairly straightforward that third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly and respect service provider decisions about whether or not to participate” [1]. Amazon added that it is in direct conversation with Meta and, asked whether it would consider legal action, declined to comment [1].

Meta launched Muse in the US earlier this month as an agent that does more than answer questions: it can shop, make payments, read email and manage calendars, and when a service offers no public API, Meta says Muse can drive it through a browser much as a person would [2]. That browser-driving approach is exactly what Amazon objects to. The Register tested the block directly, asking Muse to find the best-reviewed ergonomic office chair, add it to a cart and reach checkout; Muse reported that Amazon was “showing an anti-bot wall that blocks automated browsers outright,” and that it did not push past the notice because doing so “would violate Amazon’s terms” [2].

On the security question, Meta’s position is that Muse cannot see users’ passwords or payment methods [1][2]. The company says shared credentials go into secure storage and are made available for authentication without ever being exposed to the model, “including passwords a person types into the browser themselves” [1]. Amazon’s concern is less about whether Meta is careful and more about who is moving through customer accounts unannounced: because the agent does not identify itself, Amazon argues, an undisclosed third party can reach account pages, order history and transactions whenever a customer prompts it to [1].

What changed, and for whom

The standoff is striking because the two companies are partners, not rivals. Amazon products have been purchasable inside Facebook and Instagram since 2023, and Meta signed a deal to run agentic AI workloads on Amazon’s Graviton chips [1]. That makes this less a feud than a boundary dispute: even close partners have not agreed on the rules for an agent that acts on a customer’s behalf across a store it does not own.

It also fits a pattern. Amazon has spent much of the past year keeping outside agents off its pages — suing Perplexity over its Comet browser and moving to block shopping agents from Google and OpenAI [1]. The reason is not hard to find. Amazon’s advertising business, which depends on people browsing its pages and seeing sponsored products, took in more than $68 billion last year [1]. An agent that skips the browsing, ignores the sponsored placements and buys the cheapest match threatens the part of the shopping trip Amazon monetizes most.

For anyone who actually uses an AI assistant to shop, the practical lesson is that the agent’s abilities are only half the story. The other half is permission. An agent can be technically capable of completing a purchase and still be stopped cold by a retailer that has decided it does not want unannounced bots in its aisles. Until agents identify themselves and retailers publish terms for them, “my assistant can buy it for me” will remain conditional on where you are buying.

There is a second, quieter lesson about credentials. Even taking Meta at its word that Muse never exposes a password to the model, the fact that shared logins live in the agent’s secure storage means you are trusting a new party with the keys to accounts you already hold elsewhere. That trust is separate from whether any single retailer allows the agent through. Before you let an assistant act inside an account, it is worth asking what it retains, where that data sits, and what happens to it if you stop using the agent — the same questions you would ask any vendor you are handing a login to. Retailers blocking agents is the visible fight; the credential question is the one that follows you from store to store.

Who it is for — and not

This matters most if you have started letting an assistant do more than research — if you ask it to add items to a cart, compare prices across stores or place an order while you do something else. For you, the takeaway is to check whether the agent you use is welcome at the retailers you buy from, and to keep a manual path ready for anything it cannot finish.

It matters less if you use AI to shortlist products and then buy the winner yourself. Nothing here stops an assistant from reading reviews, summarizing specs or building a comparison; the block is specifically on an undisclosed agent transacting inside Amazon accounts. The safest posture, whichever camp you are in, is to never hand an agent a credential you would not want stored on someone else’s servers, and to prefer agents that identify themselves and use a retailer’s official interfaces rather than pretending to be a human in a browser.

This page will update if Meta responds on the record, if Amazon takes formal legal action, or if the two firms agree on terms that let Muse back in.

∴ If you shop through an AI agent, the retailer, not the agent, decides whether it may act on your behalf.