Amazon Blocks Meta's Muse AI Agent From Shopping on Amazon.com
News Summary
Amazon has cut off Meta's newly launched Muse AI agent from shopping on Amazon.com, escalating a broader industry standoff over whether third-party AI assistants can act on behalf of shoppers inside major retail platforms. Starting Sunday night, September 20, 2026, Pacific Time, Amazon began serving pop-up warnings to users who tried to complete purchases through Muse, stating that continued access by an unauthorized AI agent violates Amazon's Conditions of Use.
What Triggered the Block
Amazon said it approached Meta directly and asked the company to voluntarily exclude Amazon.com from Muse's shopping capabilities before taking any action. According to Amazon, Meta declined the request, and Amazon moved to block the agent's access late Sunday, September 20, 2026, Eastern Time. Amazon's core objection centers on three points: Meta never disclosed that Muse would browse and transact on Amazon's site, the agent does not identify itself as an automated bot while navigating pages, and it appears to capture and retain shoppers' account credentials, which Amazon says raises privacy and security concerns since it could expose order history and account details without the company's knowledge.
How Muse Works
Muse is Meta's personal AI agent, introduced on September 8, 2026, as what the company called its first agent built to handle everyday tasks such as online shopping, booking movie tickets, scheduling appointments, and filling out forms. The assistant operates through a chat-style interface, letting users name their agent and customize an avatar. Meta built Muse to run tasks inside a dedicated virtual machine hosted in its own cloud infrastructure, using a built-in browser window that is visible to the user as it completes actions. The app launched first on iOS and Android in the United States, with a free tier alongside paid subscription tiers priced at $20 and $100 per month depending on usage, and Meta has said it plans to expand availability to more countries and to its AI glasses over time.
Meta's Response
Meta has disputed Amazon's characterization of how Muse handles login credentials. The company says it relies on virtual machine isolation combined with monitoring from what it calls a Sentinel agent system to keep user passwords in secure storage without exposing raw credentials to the AI model itself. Despite the disagreement, Amazon's block has remained in effect, meaning Muse users attempting to check out on Amazon.com continue to see the warning pop-up rather than completing their purchase.
Part of a Larger Industry Pattern
This is not the first time Amazon has moved to restrict AI agents from operating on its retail platform. Over the past year, Amazon has pursued legal action against Perplexity AI over its Comet browser's shopping features and has taken steps to limit automated shopping agents built by Google and OpenAI from transacting on its site. Amazon maintains a general policy prohibiting third parties from deploying automated tools to shop on Amazon.com without authorization, arguing that unsanctioned bots can undermine the accuracy of pricing, inventory, and account security that the company guarantees to its own customers. The Muse dispute underscores a widening rift among major technology companies over who should control the checkout experience as AI agents become more capable of acting autonomously on a user's behalf.
What It Means for Shoppers
For now, consumers who want to use Muse for shopping tasks will need to complete Amazon purchases manually rather than through the agent. Meta has continued to promote Muse's broader capabilities beyond shopping, including calendar scheduling and home monitoring features, and downloads of the app have grown quickly since launch as users explore its assistant functions across categories unaffected by the Amazon restriction. Industry observers expect the standoff to continue as both companies weigh technical and policy approaches, including possible identification standards for AI agents browsing e-commerce sites, that could eventually allow supervised agent access without compromising platform security.