The Agent Wars Begin: Amazon Shuts the Door on Meta’s Muse
25 September 2026
by Paul Adams – Chief Growth Officer
Amazon has drawn a hard line in the sand of agentic commerce, blocking Meta’s new AI assistant, Muse, from shopping on its site just weeks after launch. Users who send Muse to Amazon now hit a warning that “unauthorized AI agent” access violates the retailer’s conditions of use.
Muse, launched on September 8, is Meta’s general-purpose AI agent capable of handling multistep tasks across email, calendars, dining, payments and shopping and it rocketed to the top of Apple’s US free-app chart within a week. That speed is exactly the problem for Amazon. An agent that can silently compare prices, bypass Amazon’s curated recommendations, and check out on a shopper’s behalf threatens the retailer’s core asset: control of the customer relationship and the data that fuels its personalization and advertising engine.
Amazon says Muse wasn’t authorized, doesn’t identify itself, appears to capture and store customer credentials, and scrapes account data, raising real security concerns. It’s also demanding agents be transparent, let retailers opt out, and offer “mutual exchange of value”.
This isn’t an isolated incident. Amazon sued Perplexity over its Comet browser’s shopping agent in November 2025, alleging covert, disguised access to customer accounts. A judge initially sided with Amazon in March, but a Ninth Circuit panel overturned that injunction in August, ruling users not Perplexity were the ones “accessing” Amazon. Amazon has also restricted agents from Google and OpenAI.
So why is Amazon doing this? Amazon wants agentic commerce on its own terms, not ceding checkout control to Meta, OpenAI or Perplexity’s assistants. Yet this carries commercial risk blocking ChatGPT-style agents could mean missing purchases from hundreds of millions of monthly users and Amazon’s own advertising tie-up with OpenAI suggests eventual cooperation is inevitable.
We think this is the right defensive move now, but not a sustainable end state. Amazon is protecting data integrity and security legitimately to give them time to strengthen Rufus and shape the rules of engagement but the direction of travel (its own $50B OpenAI bet, the ChatGPT ads deal and rising protocol adoption elsewhere) all points toward eventual, controlled cooperation rather than indefinite exclusion.