Amazon Blocks Meta’s Muse AI Agent Over Policy Violations

Image credit: Search Engine Journal

U.S. e-commerce giant Amazon blocked Meta‘s AI shopping agent Muse from accessing its website on Sept. 20, 2026, citing violations of its Conditions of Use and privacy concerns.

The move highlights growing tensions between major tech companies over AI agent access to proprietary platforms and reveals limitations in existing web protocols and legal frameworks designed to regulate automated access.

Amazon displayed a message indicating that Muse’s access was prohibited for three primary reasons: Meta had not informed Amazon of Muse’s activity, the agent did not identify itself as an automated system, and it appeared to capture and store customer credentials, according to technology news site GeekWire.

Traditional methods for blocking web crawlers, such as the robots.txt protocol, proved ineffective in this instance. Meta’s documentation for its web crawlers does not list Muse, and the AI agent operates as a standard Chromium-based browser session, making its activity indistinguishable from human user interactions, GeekWire’s Todd Bishop reported.

The legal grounds for Amazon’s action may be complicated by a recent court decision. The Ninth Circuit Court of Appeals ruled on Aug. 4, 2026, that the individual shopper, rather than the agent’s vendor, is considered the party accessing a website. This ruling potentially weakens Amazon’s ability to pursue claims under the Computer Fraud and Abuse Act (CFAA) against Meta.

Meta, in response, stated that Muse stores user credentials within a virtual machine (VM) unique to each individual user. The company emphasized that these credentials are not held in Meta’s centralized infrastructure. Furthermore, Meta indicated that Muse employs surrogate tokens for transactions, enhancing security.

The dispute underscores a broader challenge for websites seeking to control automated access by sophisticated AI agents that mimic human behavior. Current web protocols and legal precedents may not fully address the complexities introduced by advanced AI technologies like Muse.

Source: Search Engine Journal