Amazon Bars Meta’s Muse From Its Store as Shopify Signs On

  • AI
  • September 22, 2026
  • 0 Comments

One evening last week, a Meta customer asked the company’s new personal AI assistant to buy something on Amazon, and the order never went through. At checkout, the shopper saw a message saying an “unauthorized AI agent” was continuing to access the site in violation of Amazon’s terms of use.

The assistant is called Muse, and it had been live for less than two weeks. Meta launched it on September 8 as a personal agent that can shop, book travel, send email and fill out forms on a user’s behalf, a product the company has framed as the first consumer AI built to do work rather than answer questions.

Amazon cut Muse off on the evening of September 20, according to people familiar with the matter. The company gave three reasons: Meta never told Amazon in advance that Muse would browse its store, the agent does not identify itself as it moves through the site, and Amazon suspected Muse of scraping and storing customer login credentials.

Amazon asked Meta to remove Amazon from Muse’s experience. Meta refused. The standoff left shoppers staring at warnings in the checkout, and it drew a line between two of the largest companies in consumer technology over a question neither has answered: who is allowed to act on a customer’s behalf when the customer is not the one clicking.

The same day, Shopify took the opposite position. The e-commerce software company said it would integrate Muse across its platform for what it calls agentic checkout, letting the assistant complete purchases on the thousands of merchant sites that run on Shopify. Where Amazon saw a trespasser, Shopify saw a customer.

The split is the first real fracture in how the internet’s biggest platforms plan to handle third-party agents. For years, the dominant companies have agreed, mostly silently, that storefronts and apps belong to the people who run them. Muse, and a wave of similar agents from OpenAI, Google and others, breaks that assumption by inserting a program between a customer and a checkout page.

Muse has drawn attention at a pace that surprised even Meta’s own executives, according to analysts. Two weeks after launch, Meta’s shares climbed more than 7 percent on Monday, putting the stock on course for its best month in thirteen years. JPMorgan analysts wrote that Muse could become the most widely used consumer AI application since ChatGPT.

Mark Zuckerberg has described personal agents as the path to what he calls personal superintelligence, an always-on helper that manages a user’s health, finances, work and home. Muse is the first consumer product built around that idea at Meta’s scale, and it shipped with access through the web, iPhone and Android, and WhatsApp, with support for the company’s glasses promised later. Meta offers a free tier plus paid plans at $20 and $100 a month.

The product is built to be useful in exactly the ways that make retailers nervous. Muse fills in checkout forms, applies coupon codes and completes purchases through a single-use virtual card number generated by Stripe, so a merchant never sees the buyer’s actual card. Meta has said a user must approve a purchase before Muse completes it.

That design sits at the center of the dispute. Amazon’s terms of use have long barred automated access to its site, and the company has spent years fighting bots that scrape prices and inventory. Muse, in Amazon’s telling, behaved like one of those bots without announcing itself, and its access to customer logins made the intrusion more serious.

Ars Technica reported the same day that Muse contains a serious zero-day vulnerability, a fresh sign of the security questions that follow any agent holding a user’s passwords and payment rails. Meta did not immediately respond to requests for comment on the Amazon dispute.

For Amazon, the calculation is straightforward. It runs the largest online store in the United States, and every transaction that flows through a rival’s agent is one it does not control. Amazon has been building its own assistant, and it has little reason to let a competitor’s program mediate its relationship with shoppers.

For Shopify, the math is different. Shopify does not sell goods itself; it sells software to merchants who do. If customers begin buying through agents, Shopify’s merchants must be reachable through those agents, or they risk losing the sale to a rival. Signing on with Muse is a bet that agentic checkout will become a default way people shop, and that being early is worth the uncertainty.

The two decisions, made within hours of each other, offer the clearest picture yet of how the agent economy will sort itself out. Amazon can afford to refuse a new intermediary because it owns the customers. Shopify cannot, because it only rents the shelf.

Analysts said the episode is unlikely to be the last of its kind. As agents spread, every major platform will have to decide whether to treat them as welcome customers or as bots to be blocked, and the choices will not be uniform. What Amazon shut on a Saturday evening, Shopify had opened by the next morning.

Related Posts

  • September 24, 2026
  • 8 views
Big Cloud’s AI Bill Tops $4.2 Trillion Through 2029

For years, the largest cloud companies paid for their data centers the way they paid for everything else: out of the cash their businesses threw off. That habit is breaking.…

  • September 24, 2026
  • 9 views
Microsoft Puts $10 Billion Behind Middle East AI

Two years ago, Microsoft paid $1.5 billion for a stake in G42, an Abu Dhabi artificial-intelligence company, and took a seat on its board. This week the software giant said…