Amazon ha prevented to Meta Muse to use your own online store for make purchases on behalf of users. The measure, reported by GeekWire, appeared on Sunday in the form of a message shown to users of the AI agent: “Continued access by an unauthorized AI agent violates Amazon’s Terms of Servicewhich our customers have accepted.”
As reported, Amazon says it asked Meta to voluntarily exclude its site from the Muse experience. The problem is how the agent interacts with the store and, in particular, the fact that it would not explicitly identify itself as an automated system while browsing.
An Amazon spokesperson explained: “We find it quite clear that third-party applications that offer to make purchases on behalf of customers at other companies must operate openly and respect service providers’ decisions on whether to participate or not“.
The company adds that this approach is to ensure “a safe, secure and reliable experience” and argues that AI agents should respect the same principles already applied to, for example, food delivery platforms, delivery services and online travel agencies.
The issue does not only concern authorization to access the site. Amazon also raised concerns about its handling of user informationclaiming that Muse may access or store personal data and that it appears to acquire customer credentials while browsing.
Meta’s position is different. When Muse launched in early September, the company described the system as a “secure and private” personal agent capable of proactively helping users achieve their goals and suggesting ideas. Muse can fill out forms, send emails and carry out shopping activities, also using a single-use card generated through Stripe Links.
Meta also specified: “Muse it has no visibility into people’s passwords or payment methods. Any credentials a person shares are placed in secure storage, so Muse can use them without seeing them, including passwords a person types directly into the browser.”
The issue is therefore also linked to the difference between the agent’s technical ability to use a credential and the possibility of viewing it or storing its content. Amazon, for its part, considers the behavior of an automated system that operates on its site without identifying itself as such to be problematic.
The Muse case is part of a broader comparison between Amazon and new agentic AI platforms. The company had already taken legal action against Perplexity in November last year, challenging the AI service’s access to its site as part of Comet’s shopping experience. In August of this year, a judge ruled in Perplexity’s favor in that dispute.
Amazon has also changed some of the ways in which it communicates orders to customers in recent months. Since last July, purchase confirmation emails have become more streamlined, with the removal of elements such as specific product names and images, a change that reduces the information immediately available to automated systems that could analyze it.
The Muse case thus shows one of the problems that are emerging with the spread of AI agents: these systems do not limit themselves to providing answers, but can interact directly with external services, fill in forms and conclude transactions. It therefore becomes necessary to establish how they must identify themselves, what authorizations they must receive and what information they can process.
