Meta's Muse AI agent has been shut out of Amazon. The shopping assistant, which Meta unveiled earlier this month, can no longer place orders on the retail giant's platform on behalf of users — a restriction that surfaced without public warning and is now being communicated to users through an on-screen popup.
According to reporting by GeekWire, the notice began appearing for Muse users on Sunday. It states that "continued access by an unauthorized AI agent violates Amazon's Conditions of Use, to which our customers have agreed." The wording leaves little room for interpretation: Amazon is treating Muse not as a welcomed shopper but as an uninvited intermediary operating inside its storefront.
What Amazon is objecting to
Amazon's objections reach well past the question of who technically completes a checkout. The company says Meta never notified it that Muse would be accessing the Amazon store, meaning the assistant arrived on the platform with no prior coordination, agreement, or technical arrangement between the two firms.
Amazon also raised privacy and security concerns. In the retailer's account, Muse does not identify itself as an automated agent while it browses, and Amazon says the tool appears to be capturing customer credentials — a claim that goes to the heart of how much visibility an outside AI agent has into a user's account.
A spokesperson for Amazon framed the dispute as a matter of basic fairness among businesses. "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," the unnamed spokesperson told GeekWire.
Notably, the block is not the product of a negotiation that broke down. It is a unilateral shutdown imposed by Amazon after the fact. The Verge has reached out to Meta for comment, and the company has not yet responded publicly to the restriction.
Meta's pitch, and the privacy questions around it
When Meta launched Muse, it described an assistant designed to handle the tedious parts of online shopping while keeping sensitive data out of reach. The company said the agent cannot see secure login details or card payment information, a boundary meant to reassure users that delegating a purchase does not mean handing over their financial credentials.
In practice, the agent can apparently complete a genuine transaction. During testing, The Verge used Muse to successfully purchase tank tops from Amazon while other products sat in the basket — a demonstration of exactly the kind of automated, multi-step buying that retailers are now wrestling with.
Concerns have nevertheless accumulated since launch. Reports indicate that Muse can see the contents of user messages even though it does not have the necessary access permission enabled. That gap between what an agent promises to shield and what it can actually observe is central to why merchants are wary of letting such tools roam their systems unchecked.
There is also a structural issue Amazon is highlighting: identification. A browsing agent that presents itself as an ordinary customer is indistinguishable from a human shopper, which complicates fraud detection, rate limiting, and any policy a retailer might want to enforce against automated access.
A widening fight over who controls agentic shopping
Amazon's move against Muse is not an isolated act of gatekeeping. In November of last year, the company sued Perplexity over its Comet shopping experience, arguing that the AI search provider's agent was accessing Amazon in ways the retailer had not consented to. In August, a judge sided with Perplexity in that case — an outcome suggesting the legal footing beneath the phrase "unauthorized agent" is more contested than Amazon might prefer.
The current standoff also fits a quieter pattern of tightening controls on Amazon's side. The Verge notes that Amazon's order confirmation emails have started to look notably sparse since July, a shift that lines up with a broader effort to standardize how the retailer handles automated and third-party purchasing.
Read together, the lawsuit and the Muse block tell a consistent story: Amazon intends to be the one that decides which AI intermediaries may transact on its behalf, and on what terms. The company's interest is easy to trace — every purchase an external agent mediates is a purchase where Amazon loses control of the customer relationship, the merchandising surface, and the data that comes with it.
Why identity is the sticking point
At the center of this dispute is a question the industry has not answered: what does it mean for an AI agent to act "openly"? For Amazon, it means declaring itself, accepting the platform's terms, and abiding by a merchant's decision to opt out. For agent developers, the same demand can read as a requirement to negotiate access with every retailer on the web before a user's own shopping assistant can do what that user asked it to do.
- Whether Meta pursues a formal integration agreement with Amazon or leaves Muse blocked.
- Whether other large retailers adopt Amazon's approach and begin screening for automated agents.
- How courts interpret existing conditions-of-use language when the "user" is software acting on a person's instructions.
- What recourse shoppers have when a tool they paid for or adopted loses access to a major merchant overnight.
- Whether delegated authentication standards emerge that let agents identify themselves without exposing credentials.
The unresolved nature of those questions is precisely why the Muse block matters beyond two tech giants trading barbs. Agentic commerce only works if agents can reach the places where people actually buy things — and right now, the largest of those places has drawn a line.
What happens next
For the moment, Muse users who leaned on the assistant to buy from Amazon will have to finish those purchases themselves. The popup gives no timeline for restoration, and there is no indication that Meta and Amazon are in talks to resolve the impasse.
Meta's options are limited but real. It could seek a formal commercial or technical agreement with Amazon, modify Muse to announce itself as an automated agent, or route its users back to Meta-operated surfaces where it controls the terms. Each path carries trade-offs in capability, cost, and user experience.
Amazon, meanwhile, has signaled that it would rather absorb short-term friction than cede the checkout page. The company has already spent a year litigating a version of this argument with Perplexity, and its position has not softened despite the August ruling. Expect the retailer to keep treating unidentified agents as a category to be excluded rather than accommodated.
The larger consequence is that the agentic shopping era will not be built purely on better models. It will be built — or blocked — at the level of access. Whoever decides which agents may transact, and under what disclosures, effectively sets the rules for how AI assistants participate in commerce at all. For now, Amazon has made its answer clear, and Meta's Muse is on the wrong side of it.
This article is based on reporting by The Verge. Read the original article.
Originally published on theverge.com
