Amazon is blocking Meta’s Muse from accessing its marketplace, exposing an emerging fault line in the race to build AI agents that can shop, transact, and carry out tasks on behalf of users.
People using Meta’s personal AI assistant to access Amazon are encountering a warning that “continued access by an unauthorized AI agent” violates the retailer’s conditions of use.
Amazon said it had asked Meta to remove its marketplace from Muse, arguing that third-party AI agents cannot assume access to commercial websites simply because users want them to perform tasks there.
“Third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly and respect service provider decisions about whether or not to participate,” an Amazon spokesperson said.
“Agentic third-party applications such as Muse have the same obligations,” the spokesperson added.
The dispute highlights a problem that could become more valuable as AI agents move beyond answering questions and begin acting independently across the internet. A chatbot can provide a link to a product. An agent can potentially search for the product, compare prices, log into an account, and complete the purchase.
That difference gives retailers a new reason to control how AI systems interact with their platforms. Amazon says Muse was not authorized to access its store, does not identify itself as an AI agent, appears to capture and store customer credentials, and scrapes account data. The company also argues that automated agents can bypass elements of its personalized shopping experience, potentially presenting products without taking into account a customer’s purchase history and recommendations.
Amazon said it expects AI agents to be transparent about their activity, give retailers the ability to opt out, and establish a “mutual exchange of value” with the businesses whose platforms they access.
The confrontation was flagged by Muse users on X after Amazon’s bot-detection systems began blocking access. Startup founder Jonathan Wegener posted a screenshot of the warning, highlighting the direct conflict between AI agents and websites designed around human users.
Amazon’s existing conditions of use do not explicitly refer to AI shopping assistants, but they prohibit the use of “data mining, robots, or similar data gathering and extraction tools.” The company has relied on that language as it seeks to limit automated access to its marketplace.
The dispute with Meta is also not an isolated case. Amazon has sued Perplexity over shopping activity conducted through its Comet browser and has moved to restrict agents from Google and OpenAI.
Meta launched Muse on September 8 as a general-purpose AI assistant capable of handling multistep tasks involving email, calendars, dining, payments and shopping. The application reached the top of Apple’s US free-app chart within a week, underscoring rapid consumer interest in systems that can do more than generate text or answer questions.
The Amazon dispute, however, exposes the limits of that model.
AI companies can build agents capable of navigating websites, but they do not necessarily control whether those websites will accept automated activity. As agents become more autonomous, access rights, authentication, data handling, and commercial agreements could become as important as the underlying AI capabilities.
The issue is especially sensitive in e-commerce because the agent sits between the consumer and the retailer. An agent that chooses products, decides where to buy them and completes transactions could influence which merchants receive sales, how products are ranked and what customer information is shared.
For retailers, that creates both an opportunity and a threat. AI agents could generate additional purchases by making shopping easier, but they could also weaken the direct relationship between retailers and customers. An intermediary agent could determine what consumers see, which products they compare, and where their money ultimately goes.
Amazon has already indicated that it wants a commercial framework for that relationship rather than unrestricted access.
There is also a potential contradiction in the emerging AI shopping market. Agents need access to large retail platforms to become useful, while those platforms have little incentive to surrender control of customer data, recommendations, and transactions without receiving something in return.
That tension could force the development of a new layer of agreements between AI companies and online businesses. Instead of agents treating the open web as an unrestricted operating environment, retailers may demand authenticated access, explicit identification of AI systems, limits on data collection and commercial arrangements governing transactions.
The economics could become significant as well. If AI agents become a major channel for online shopping, retailers will have to decide whether to treat them as another source of traffic, paid distribution partners, or competitors for the customer relationship.
Amazon’s relationship with OpenAI adds another complication. The companies recently announced an advertising partnership, suggesting that Amazon has a commercial incentive to work with at least some AI systems even as it resists unauthorized automated shopping activity.
For Meta, the challenge is more fundamental. Muse’s value depends partly on its ability to act across services that Meta does not control. If major platforms can independently block the agent, its ability to function as a general-purpose assistant becomes dependent on a patchwork of permissions and commercial agreements.
The Amazon dispute therefore goes beyond a single bot blocker. Many see it as an early test of who controls the next interface to the internet: the platforms that own the websites and customer relationships, or AI agents that want to act on behalf of users across those platforms.
Muse may be ready to shop, but Amazon’s response shows that retailers are not necessarily ready to let AI agents do the shopping on their terms.
