Sunday night, if you tried to buy something on Amazon through Meta’s new AI assistant, you got a message instead of a checkout screen: “Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use.” Just like that, Amazon shut the door on Muse, and the fight over who gets to control how AI agents shop on the internet got a lot more public.
What Muse Actually Is
Meta launched Muse on September 8, and it’s a more ambitious product than most people probably realized at the time. It’s not a chatbot that answers shopping questions — it’s a personal agent designed to handle multi-step tasks across email, calendar, payments, restaurant bookings, and shopping, all on your behalf. It runs on iOS, Android, through muse.ai directly, and inside WhatsApp, and it operates in a sandboxed virtual machine with its own browser rather than borrowing your device’s session. Meta built in a guardrail requiring explicit user approval before anything sensitive happens, purchases included.
On paper, that sounds like a reasonably careful approach to a genuinely useful idea: an assistant that can actually go do things instead of just describing how you’d do them yourself. Amazon apparently didn’t see it that way.
Why Amazon Pulled the Plug
Amazon’s case against Muse comes down to three specific complaints. First, the company says the agent doesn’t identify itself when it’s browsing Amazon’s site, making it indistinguishable from a regular human shopper at the traffic level. Second, Amazon claims Muse appears to capture and store customer credentials in a way that creates real privacy and security exposure. Third, and maybe most pointed, Amazon says Meta never disclosed that Muse would be operating on its platform in the first place and never asked permission to build the integration.
An Amazon spokesperson put the company’s position plainly: third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly and respect service provider decisions. Amazon has reportedly asked Meta directly to pull Amazon out of Muse’s shopping capabilities, drawing a comparison to how food delivery apps and travel agencies operate — with explicit merchant agreements, not by quietly showing up and browsing like a customer.
Meta’s Side of the Story
Meta has pushed back on the security framing specifically, stating previously that Muse “has no visibility into people’s passwords or payment methods,” and that credentials are stored in a way the agent itself can’t directly access. That’s a direct rebuttal to one of Amazon’s three core claims, though it doesn’t really address the identification or permission issues Amazon also raised. As of this weekend, Meta hadn’t issued a broader public response to the block itself.
This Isn’t Amazon’s First Fight Over Agentic Shopping
If this feels familiar, that’s because it is. Amazon has spent the better part of the past year trying to keep outside AI agents off its marketplace, and Meta is just the latest company to run into that wall. Perplexity got there first, with its Comet browser’s shopping agent. Amazon sued, won a preliminary injunction back in March, and then lost it in August when the Ninth Circuit ruled that under federal anti-hacking law, it’s users — not the AI companies building the tools — who are technically the ones accessing the website. Amazon asked for a rehearing on that ruling; it was denied on September 10, just eleven days before the Muse block.
Google and OpenAI have both had their own shopping agents blocked by Amazon at various points too. This is a pattern now, not an isolated dispute, and the Muse situation is really just the newest chapter in a fight Amazon has been fully committed to fighting on every front simultaneously.
What Amazon Wants Instead
It’s worth noting Amazon isn’t opposed to AI shopping agents in general — it just wants to be the one running them. The company launched its own Alexa for Shopping feature back in May, along with a “Buy for Me” tool, both of which identify themselves clearly when active and give users the ability to opt out. That’s a meaningfully different approach than what Amazon is accusing outside agents of doing, and it’s also, not coincidentally, an approach that keeps Amazon fully in control of the shopping experience on its own platform.
The Money Behind the Fight
There’s a very concrete reason Amazon is fighting this hard, and it’s not really about security theater. Amazon pulled in more than $68 billion in advertising revenue last year, and that entire business model depends on people actually browsing Amazon’s site, seeing sponsored listings, and making decisions influenced by what Amazon chooses to show them. An AI agent that skips straight to a purchase decision, bypassing the browsing experience entirely, threatens that revenue stream directly. Every company building an agentic shopping tool is, whether they frame it this way or not, also building a tool that could route around retailers’ ad businesses. That’s the real fight underneath the security and disclosure arguments.
Where This Leaves Consumers
For now, anyone who set up Muse expecting it to handle Amazon purchases is out of luck. The agent presumably still works for email, calendar management, and dining reservations, along with whatever other retailers haven’t blocked it yet. But the core promise of agentic shopping — an assistant that seamlessly handles purchases across any store you want — is running headfirst into the reality that the biggest store on the internet has no intention of making that easy for anyone it hasn’t personally authorized.
What This Means
The bigger story here isn’t really Amazon versus Meta specifically. It’s that the entire agentic AI shopping category is currently being decided in legal filings and platform-level blocks rather than through any kind of negotiated standard. Every major AI company wants an agent that can shop for you, and every major retailer wants to control exactly how, when, and whether that happens on their turf. Until there’s some kind of industry framework — an agreed-upon way for agents to identify themselves, request access, and share revenue with the platforms they’re shopping on — expect more of exactly this: agents getting blocked, companies trading public statements, and courts occasionally getting pulled in to sort out who technically owns the click. Muse won’t be the last agent Amazon shuts out, and Amazon almost certainly won’t be the last platform doing the shutting out.
