Why Amazon Keeps Blocking AI Shopping Agents

Photo by Igor Omilaev on Unsplash

Today Amazon cut off Meta's Muse from shopping on Amazon.com. If you try to send the agent to buy something there now, you get an error message citing a violation of Amazon's Conditions of Use. It's a small technical move with a much bigger story behind it: the fight over who actually controls a shopper's path from "I want this" to "it's on its way."

What happened

Meta launched Muse in the US on September 8. It's a personal AI agent that goes beyond answering questions. It can shop, read your email, manage your calendar, and make payments on your behalf, and when a service doesn't offer a public API, it just operates a browser the way a person would, clicking and typing its way through a site.

Muse caught on fast. Within a week it was the top free app on Apple's US App Store, ahead of ChatGPT. Then, on September 21, Amazon shut the door. The error message Muse now hits on Amazon.com says access by an "unauthorized AI agent" violates the site's Conditions of Use.

Amazon's stated reasons come down to three things. First, Muse doesn't identify itself while it's browsing, so from Amazon's side it looks like ordinary human traffic rather than a bot. Second, Amazon says the agent captures and stores customer credentials, which it considers a security risk. Third, Amazon argues agents should get a merchant's consent before acting on a site, the way food delivery apps partner directly with the restaurants they serve rather than just scraping the menu.

Meta pushed back on the security claim specifically. A spokesperson said Muse "has no visibility into people's passwords or payment methods," and that credentials are held in encrypted storage and used for authentication without ever being exposed to the model itself. Meta hasn't responded point by point to Amazon's other objections.

There's a less flattering read of Amazon's motives sitting right next to the security argument, too. Amazon pulled in more than $68 billion in advertising revenue last year, and that business depends on shoppers discovering products through Amazon's own search and recommendation systems. An agent that skips straight to a purchase, bypassing the sponsored listings and the browsing behavior Amazon's ad engine is built on, is a threat to that revenue no matter how well-behaved it is.

This isn't a one-off fight

Muse is just the latest agent Amazon has shut out. Amazon sued Perplexity over its Comet browser agent and won a preliminary injunction back in March, though the Ninth Circuit reversed that ruling in August, deciding that users, not the AI companies building the tools, are the ones who get to decide who accesses a site on their behalf. Amazon has also blocked shopping agents built by Google and OpenAI.

Amazon isn't just playing defense, either. It runs its own agents, Alexa for Shopping and Buy for Me, and points out that Buy for Me identifies itself to retailers and lets them opt out if they don't want the traffic. That's a real difference from how Muse currently operates, and it's the crux of Amazon's complaint: not that agents shouldn't be allowed to shop, but that they should show up announced.

And despite the standoff, Amazon and Meta aren't exactly enemies. The two companies signed a multibillion-dollar cloud deal back in April, with Meta using Amazon's infrastructure. The shopping block sits alongside that partnership rather than replacing it, which tells you this is a turf dispute over one specific piece of the relationship, not a full-blown corporate war.

The bigger pattern: agentic commerce is running into walls

What's playing out between Amazon and Meta is a preview of a problem every retailer is going to face as AI agents get better at acting on the open web. An agent that browses like a human, without announcing itself, puts a site in an awkward spot. Site owners can't tell real customers from bots doing the customers' bidding, they can't verify what the agent is actually doing with stored credentials, and they lose the ability to shape the experience, whether that's cross-selling, loyalty programs, or simply showing the ads that pay for the site in the first place.

The standards that would fix this don't really exist yet. There's no agreed way for an agent to identify itself, no common protocol for a merchant to grant or revoke access, and no shared definition of what "authorized" even means across different platforms. Until something like that emerges, expect more of this pattern: an agent launches, it gets popular fast because it's genuinely useful, and then the sites it depends on start locking the door behind it.

For developers, this is worth paying attention to beyond the Amazon-Meta drama. If you're building anything that touches agentic commerce, whether that's an agent that shops or a storefront that wants to be agent-friendly, the lesson from this dispute is that identification and consent aren't optional extras. They're going to be the actual battleground, probably before pricing or capability ever is.

Key takeaways

Amazon blocked Meta's Muse from shopping on Amazon.com on September 21, citing unauthorized access, hidden identity while browsing, and credential storage concerns. Meta disputes the credential claim, saying passwords and payment details are never exposed to the model. This is part of a broader pattern: Amazon has previously restricted or sued over shopping agents from Perplexity, Google, and OpenAI, even as it runs its own agents, Alexa for Shopping and Buy for Me. The underlying issue for the whole industry is that there's no shared standard yet for how an AI agent should identify itself and get a merchant's consent before acting on their site.

FAQ

What is Meta's Muse? Muse is Meta's personal AI agent, launched in the US on September 8, 2026. It can complete multi-step tasks like shopping, managing email, and handling payments, using a browser directly when a service has no API for it to call.

Why did Amazon block Muse? Amazon says Muse accessed its site without authorization, doesn't identify itself as an AI agent while browsing, and stores customer credentials in a way Amazon considers a security risk. Amazon also wants agents to operate with merchant consent rather than browsing unannounced.

Has this happened to other AI agents? Yes. Amazon previously sued Perplexity over its Comet browsing agent and won an injunction that was later overturned on appeal, and it has also blocked shopping agents from Google and OpenAI.

Does Amazon have its own AI shopping agent? Yes, two: Alexa for Shopping and Buy for Me. Amazon notes that Buy for Me identifies itself to retailers, who can choose to opt out, which is the behavior it wants from third-party agents too.