Amazon has blocked Meta’s Muse AI assistant from making purchases on its marketplace, telling users that access by an unauthorized AI agent violates its Conditions of Use. The restriction surfaced Sunday night when people attempting to shop through Muse began receiving an error message instead of completing purchases.
The block creates an early dispute over how autonomous shopping agents should interact with online stores they do not control. AI agents increasingly promise to search for products, compare options and complete purchases on a user’s behalf, raising questions about whether retailers must allow outside software to interact with their stores in the same way human customers do.
Amazon Does Not Want Muse Shopping on Its Store
Users attempting to send Muse to Amazon were shown a warning stating that continued access by an unauthorized AI agent violated Amazon’s rules, according to GeekWire. Anyone using Muse to purchase an Amazon product must therefore complete the transaction another way or find the item through another retailer.
The dispute comes as major technology companies develop competing approaches to AI agents and online commerce. Amazon has its own AI infrastructure business, including its Nova foundation models and Bedrock platform, while Meta is trying to make Muse capable of performing tasks across services on behalf of users.
Amazon has not been legally required to give Muse access to its marketplace. The decision therefore raises a larger question for agentic commerce: whether an AI acting on a customer’s instructions should be treated like the customer themselves or as a separate automated service that needs permission from the retailer.
Shopping Agents Introduce New Responsibility Questions
There are also practical reasons retailers may be cautious about allowing third-party agents to complete transactions. If an AI selects the wrong product, misunderstands quantities or completes a purchase the user did not intend, the retailer could still have to handle returns, complaints and disputes involving customers and sellers.
Those risks become more significant because current AI models can still generate incorrect information. Muse performs relatively well on some hallucination evaluations, but its error rate is not zero, leaving room for mistakes when an agent moves from recommending products to spending a user’s money.
Amazon’s decision does not necessarily mean it opposes agent-driven shopping altogether. The company is already incorporating generative and agentic AI into its own shopping experience, where it can control how the software interacts with customer accounts, sellers and transactions.
The Muse block instead illustrates an emerging issue as AI assistants become capable of acting across the web: agents may technically be able to use a service without that service agreeing to participate.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.