Amazon blocks Meta's Muse AI agent from shopping its store

By VK

Published 21 Sep 2026, 13:17 UTC

Reporting from The Verge, TechCrunch

Amazon has started blocking Meta's Muse agent from shopping for users, saying it breaks Amazon's conditions of use and citing privacy and security concerns, since Muse does not identify itself while browsing. Amazon previously sued Perplexity over its Comet shopping agent.

What it means for founders

Platform access is not guaranteed: An agent that works on a large site today can be shut off overnight through a terms-of-use change or a popup, even after a public launch. Plan for key destinations saying no.
Identify your agent: Amazon's complaints center on Muse not declaring itself and on credential handling. Agents that announce themselves, keep credentials out of reach and seek permission are harder to block on these grounds.
Courts are not settling it fast: A judge sided with Perplexity in August, yet Amazon is still blocking agents by other means. Expect retailers to use technical and contractual barriers regardless of how lawsuits go.
Watch the data trail: Amazon stripping detail from confirmation emails shows retailers may also restrict the indirect data agents rely on, not just direct browsing.

The story

Meta's Muse assistant can no longer buy things on Amazon for the people who use it. Starting Sunday, Muse users who tried to shop on Amazon.com were stopped by a popup telling them that an AI agent Amazon had not authorized was breaking the store's Conditions of Use, terms its customers accept. GeekWire first spotted the change, which was then picked up by The Verge and TechCrunch.

Amazon's stated reasons

According to GeekWire's reporting, as relayed by The Verge, Meta never told Amazon that Muse would be accessing its store. GeekWire also reported that Amazon voiced privacy and security concerns about the way Muse browses without identifying itself as an agent and about it apparently capturing customers' credentials.

An unnamed Amazon spokesperson framed the issue as one of openness. In Amazon's view, any third-party app that buys from another business on a customer's behalf should be transparent about what it is and should accept that business's choice about whether to take part.

The Verge said it had asked Meta for comment. Neither outlet carried a response from Meta.

What Meta has said about Muse

Muse launched earlier this month. At that time, Meta said the agent had no view of users' protected sign-in data or their card numbers when paying. The Verge had used Muse in testing to complete an Amazon purchase of tank tops, even with other items sitting in the basket. Since the launch, though, some privacy questions have surfaced, including reports that Muse can read the contents of users' messages without the relevant permission being switched on.

A pattern of keeping agents out

This is not Amazon's first move against third-party shopping agents. In November last year it sued Perplexity, trying to keep its store out of the shopping features in Perplexity's Comet browser. A judge ruled in Perplexity's favor in August. Separately, The Verge noted that Amazon's order confirmation emails have become much sparser since July, dropping item names and product photos, a change that limits how much outside AI services can extract from a customer's inbox.

TechCrunch offered two readings of the block. One is competitive: Amazon runs its own family of foundation models and one of the most widely used inference platforms, so it has little reason to open its storefront to a rival's agent unless the law requires it. The other is operational. When an agent places a wrong order, Amazon is left handling the fallout with both the unhappy buyer and the seller. Even if Amazon sees promise in agent-driven shopping, TechCrunch suggested, it may prefer to wait until the technology matures further.

For Muse users, the practical result is simple: purchases through the agent have to happen at some other retailer.