Amazon blocks AI shopping agents. The real question: who is the agent?
Amazon has started to block AI agents that shop on behalf of customers. Elon Musk has reacted publicly. Many people see this as a fight between big companies. For everyone else, it points to a simpler question that a ban does not answer: who is the agent, and who stands behind it?
What happened
- Around 20 September 2026, Amazon blocked Muse, the AI shopping agent of Meta. Amazon says its Conditions of Use require agents to identify themselves as agents in every request, and that Muse does not do so and does not let brands opt out. Amazon's own agent, Buy for Me, was not blocked; Amazon says it identifies itself and lets brands opt out. GeekWire
- Amazon is also in court with Perplexity over an agent that shopped on Amazon through customers' own accounts. The case has been running since November 2025 and is not settled. GeekWire
- On 15 October 2026, Elon Musk wrote on X, in reply to a discussion about Amazon restricting AI bots from placing orders: "Maybe we should start a competitor to Amazon." Reports that Amazon also blocked Grok come from users on X; we have not found a confirmation from Amazon. American Bazaar
Why a blanket ban does not settle it
Three parties have a legitimate interest, and they pull in different directions.
- The customer wants to hand a task to an assistant: compare, reorder, return.
- The shop wants to know who is acting on its site, with whose permission, and wants control over its brands and customer relationship.
- The operator of the agent needs rules it can follow, and a way to show that it does.
A ban protects the shop and leaves the other two without an answer. “Let everyone in” does the opposite. What all three need is the same thing: a way to say who an agent is and what it may do, in a form that others can check.
“Identify yourself” has two layers
- Technical marking of each request. This is what Amazon asks for in its terms: the agent says in every request that it is an agent. That is up to the agent's software and the shop's rules. A registry does not replace it.
- Who stands behind the agent, and what may it do? A request marked “I am an agent” still does not tell you which company is behind it, whether that company exists, or what the agent is authorised to do. This is the question ID MANDAT answers.
What ID MANDAT adds
ID MANDAT is an open registry. A company declares on its own domain which AI agents act for it and what each agent may and may not do. Anyone can check this, without an account:
- by web:
idmandat.com/en/check.php?domain=example.com - by program:
https://idmandat.com/api/v1/verify?domain=example.com, with a digitally signed answer (Ed25519, beta) - by AI agent: through an MCP server and an A2A endpoint
An answer says whether the domain is registered, at which level, which agents exist, and what they may and may not do. If an entry is revoked or suspended, the answer says so, so a shop or another agent does not have to guess.
What this does not do: it does not decide whether a shop must accept agents, and it is no guarantee that any shop will. The details on agents and their permissions come from the company itself. ID MANDAT confirms identity and authority, not creditworthiness, trustworthiness or quality.
What follows from this
If agents are going to buy, book and write for companies and customers, the question “who is this?” will come up thousands of times a day. It is better answered once, publicly and checkably, than separately inside every platform. That is a position, not a prediction: we do not know how shops and agent makers will settle their dispute.
Register your company · Check a domain · Technical standard · What is ID MANDAT?
Sources
- GeekWire, “Amazon's fight with Meta: who owns the customer relationship when an agent does the buying” (September 2026)
- American Bazaar, “Elon Musk hints at launching Amazon competitor” (15 October 2026)
For shops: how to ask who is behind a signed agent, and how to set house rules for agents.