Amazon kicks Meta's AI shopper off its store

Meta's new personal AI assistant Muse hit the top of the App Store within days of launch. But when it tried to go shopping on Amazon.com, it ran straight into a digital brick wall.

Over the weekend, users attempting to use Muse to buy items on Amazon were greeted with an error stating that continued access by an unauthorized AI agent violates Amazon's Conditions of Use. The Register tested the block by asking Muse to find and purchase a chair, only for the bot to get stopped before it could even load search results.

Amazon says it pulled the plug after Meta refused a private request to voluntarily exclude Amazon from the service. The e-commerce giant claims Muse browses without identifying itself, accesses user order history without merchant consent, and appears to store credentials.

The bot that pretends to be you

At the center of the dispute is how Muse handles the web. Launched on September 8, Muse is designed to execute multi-step chores rather than just answer prompts. According to Meta, if a service lacks a public API, Muse uses an isolated cloud browser to interact with the site just like a human would.

Meta has insisted that Muse has zero visibility into passwords or payment methods, routing shared credentials into secure storage and running transactions past a dedicated monitor agent called Sentinel. Sensitive actions also require final user approval.

Amazon sees it differently. A bot masquerading as a regular shopper without declaring itself creates obvious security blind spots. Amazon drew parallels to food delivery apps or online travel agencies, which operate through formal commercial agreements with restaurants and airlines rather than scraping behind the scenes.

The standoff is particularly awkward given the companies' tight ties. Amazon products have been sold natively inside Instagram since 2023, and Meta signed a multi-billion dollar deal in April to run agent workloads on AWS Graviton chips.

Fighting over a $68 billion ad machine

While Amazon frames this around consumer safety, the financial incentives are hard to ignore. Amazon brought in over $68.6 billion in advertising services revenue last year, driven heavily by sponsored product listings.

If millions of shoppers dispatch autonomous agents to pick items and bypass search result pages, sponsored placements lose their audience instantly. Amazon keeps the transaction, but loses the high-margin ad impression that sellers paid for.

Amazon is also battling on the legal front. After suing Perplexity last year over its Comet browser, Amazon won an injunction that was subsequently vacated on August 4 by the Ninth Circuit, which ruled that users accessing computers via an agent did not violate federal anti-hacking laws. Shut out of that legal avenue, Amazon is now leaning hard on its Terms of Service.

Meanwhile, Amazon already operates its own AI retail tools, including Alexa for Shopping and a Buy for Me feature. Crucially, Amazon notes that its own agent identifies itself to external stores and offers an opt-out mechanism.

Why it matters

For consumers, this spat is an early reality check on the promise of autonomous agents. Muse shot to No. 1 on the App Store because delegating repetitive online errands feels genuinely magical. Having that loop broken on the world's largest store pushes users back to manual browsing or fragmented retailer apps.

For the AI sector, Amazon is establishing a blueprint for defensive platform moats. We think agent providers will soon discover that pretending to be a human in a headless browser is an unsustainable product strategy. When an agent threatens high-margin ad models, merchants will simply rewrite terms of service and block traffic at the firewall.

In our view, agentic commerce will not scale through brute-force web automation. It will require commercial pacts, shared affiliate revenue, or ad-supported agent protocols. Until Meta and Amazon cut a deal, Muse's retail ambitions will stop at Amazon's front door.