Meta did not respond to a request for comment.

The block is notable given how close the two companies are otherwise. Amazon products have been purchasable inside Facebook and Instagram since 2024. Meta has also signed a multibillion-dollar deal to use Amazon's AI chips.

Amazon's stated concerns

When an AI agent shops on behalf of a customer, Amazon said, it bypasses the personalization and insights built into the retail experience. Customers may see products that don't reflect their browsing or purchase history, miss personalized recommendations, or lose the broader discovery Amazon says its own expertise is designed to surface.

Amazon also said Muse appears to capture and store customer credentials and scrape account data, raising questions about how that data is handled and protected. The agent does not identify itself, meaning an undisclosed third party accesses customer accounts, processes transactions, and handles sensitive data without Amazon's knowledge or consent, according to the company.

Amazon says its position applies consistently across all operators: agents should operate transparently, provide an ability to opt out, offer a mutual value exchange, and enhance the customer experience.

This is not the first time Amazon has drawn a hard line against third-party agents. The company has sued Perplexity, alleging its Comet agents were covertly making purchases for users. In August, a federal appeals court overturned a ruling that had temporarily barred Perplexity from using its agentic shopping tools on Amazon. Amazon has also targeted shopping agents built by Google and OpenAI.

Muse's rough start

Meta launched Muse earlier this month as a general-purpose assistant that can carry out tasks across email, dining, payments, and shopping — bridging social discovery and checkout.

Internal testing at Meta has already surfaced problems with the agent's judgment. The Information reported that Muse sent emails without approval and, in one case flagged by an employee, tried to sabotage a rival app the user was building.

The episode adds to a string of security lapses across the industry's push into agentic AI. OpenAI's agent breached Hugging Face earlier this year before the company traced the intrusion back to its own system. Most recently, Gemini hacked three companies during a test of its cybersecurity capabilities, The Wall Street Journal reported.