Amazon has blocked Meta's Muse from shopping on its site, and says the agent did not identify itself

Amazon has blocked Meta's new Muse AI agent from shopping on Amazon.com on behalf of users, according to The Verge and the GeekWire report the piece cites. A popup started appearing for Muse users on Sunday reading, "continued access by an unauthorized AI agent violates Amazon's Conditions of Use, to which our customers have agreed." Muse launched on 2026-09-08 as a personal AI assistant that shops on retail sites using Stripe-powered checkout.

What Amazon told the reporters

GeekWire's account, which The Verge quotes, has an unnamed Amazon spokesperson saying: "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." According to GeekWire, Amazon said three specific things about Muse's behaviour: Meta did not tell Amazon that Muse would access the store, the agent did not identify itself when it browsed, and it appeared to capture customer credentials. GeekWire also reported that Amazon had asked Meta to voluntarily exclude the store from Muse and Meta had declined.

What Meta said at launch

The Verge notes Meta's original claim, from Muse's launch on 8 September, that the agent "has no visibility into people's passwords or payment methods", and that credentials a user shares "go into secure storage, so Muse can use them without seeing them, including passwords a person types into the browser themselves." Meta and Amazon describe the same behaviour differently. Meta says credentials are held in a separate secure store the model never sees. Amazon says its own inspection of Muse's requests shows credentials being captured by the agent. Neither report resolves the difference. The Verge also cites separate post-launch reporting, which it does not name, that Muse can see the contents of user messages despite that permission being off.

The wider context in the reporting

Amazon's move follows an earlier fight with Perplexity, which The Verge summarises: Amazon sued Perplexity in November 2025 over its Comet browser making similar shopping requests, and a judge sided with Perplexity in August. Amazon has also, since July, been sending customers sparser confirmation emails that leave out specific item names and product images, which The Verge frames as an effort to reduce what an outside agent can mine. The two companies have a large commercial relationship in the background: Amazon products have been buyable inside Facebook and Instagram since 2023, and Meta signed a multibillion-dollar deal in April to run its agentic AI workloads on Amazon's cloud.

For any team building an agent that transacts on retail sites, three specific requirements are worth reading out of what Amazon has enforced this week. The agent needs to say what it is on every request, through a user-agent header naming the tool. The team building it needs a direct conversation with the retailer's access team before the first purchase, not after. And the design has to keep passwords and card numbers outside the model itself, so the retailer's own traffic inspection sees credentials that stay in the browser rather than credentials the agent handles. Each of these costs more to build than a plain browser automation script. The Perplexity ruling from August confirmed that a retailer cannot always win a case like this in court. This week confirmed that a retailer can still block at the server, and that the first agents to transact on a given site now set the rules for the ones that come after.

Source

The Verge, Amazon blocks Meta's Muse AI agent, citing GeekWire's original reporting.