Amazon’s decision to block Meta’s Muse from shopping on Amazon.com raises an important question about agentic commerce: if an AI arrives acting for a person, does the retailer have to let it in?
GeekWire reported that Amazon cut off Meta’s new Muse personal AI agent from shopping on Amazon.com on behalf of customers after trying unsuccessfully to get Meta to exclude the retailer from the experience. People using Muse to shop on Amazon began seeing a warning that described the tool as an unauthorized AI agent and said continued access violated Amazon’s Conditions of Use.
Amazon’s position is that Meta did not tell the company Muse would access its store, that Muse does not identify itself while browsing, and that the agent appears to handle customer credentials in ways Amazon did not approve. Amazon told GeekWire and Retail Dive that third-party applications making purchases from other businesses should operate openly and respect a service provider’s decision about whether to participate.
That puts Amazon’s argument on permission, not simply performance. The issue is not only whether Muse can find a product or complete a purchase. It is whether Amazon accepts a third-party AI agent moving through customer accounts, scraping or processing data, and handling transactions inside Amazon’s retail environment.
Meta designed Muse to act, not just answer
Meta introduced Muse on September 8 as a personal AI agent built to complete tasks rather than only respond to prompts. Meta says Muse runs inside a dedicated secure virtual machine, can work across apps and services, and asks for approval before higher-risk actions such as sending an email or making a purchase.
Meta’s launch materials describe Muse as available in the U.S. on iOS, Android, and the web, with WhatsApp integration and future plans for AI glasses. The company also said Muse can check out with Link by Stripe and that Shop Pay support is planned.
That design is exactly what makes the dispute hard. A normal chatbot answers a question and stops. An agent takes instructions, enters other services, fills forms, compares options, and can get close to moving money. Once that happens, every party in the transaction starts caring about identity, authority, records, liability, fraud, refunds, and customer data.
A patched bug is easier than an access policy
The Amazon block landed during the same week VentureBeat reported that Meta patched a Muse zero-day vulnerability in the Mac app after a disclosure by security researcher Patrick Wardle. That security issue matters, especially because personal agents can connect to accounts and act with the access their users provide.
But a software flaw can be patched. An access dispute is harder. It requires agreement over whether an AI agent is more like a browser, a bot, a customer representative, a payment instrument, or a new kind of actor that needs its own rules.
Amazon has made clear that it is not opposed to shopping AI in general. Its own retail AI work includes Rufus, which Amazon describes as a shopping assistant that can answer shopping questions, provide personalized recommendations, track prices, and shop agentically on a customer’s behalf. The tension is about outside agents entering Amazon’s store without Amazon’s approval.
The web was built around humans and bots
For years, online services have tried to sort traffic into a simple pair of categories: human visitors and automated bots. Humans were generally welcomed. Bots were challenged, rate-limited, blocked, or sent to APIs with rules attached.
AI shopping agents blur that distinction. A tool can be automated and still be acting for a real customer with real money and a real intention to buy. Visa made a similar point in recent agentic-commerce commentary, saying merchants have spent years identifying bots, fraud, and suspicious behavior, while legitimate AI agents may look similar to threats under older systems.
Tech Help Canada has covered how AI is changing search from link lists into answers and comparisons. Shopping agents push the same shift into checkout. The agent may choose the product, compare options, and take action before a person ever lands on a retailer’s product page.
Banks are raising the same concerns
The Amazon-Meta dispute is not happening in isolation. On September 22, ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia, ING Group, and NatWest Group announced shared principles for trusted agentic commerce. The banks said the principles focus on transparency, safety, privacy and data, choice, and interoperability.
Bank of America’s release said customers and merchants should retain choice, control, and flexibility in how payments happen while keeping transactions safe and secure. Reuters also reported that banks warned AI shopping bots could raise scam, fraud, and data-privacy risks.
Those concerns map directly onto the Amazon-Meta standoff. If an agent buys something, the payment system needs to know what the customer authorized, what the agent actually did, which merchant received the order, and who is responsible if the order, refund, fraud claim, or data handling goes wrong.
The legal line is still forming
Amazon has already tested related arguments in court. In Amazon.com Services v. Perplexity AI, Amazon sued over Perplexity’s Comet browser and its AI assistant. The Ninth Circuit vacated a preliminary injunction on August 4, 2026, finding that Amazon was unlikely to succeed, on the record before the court, in showing that Perplexity itself accessed Amazon’s computers for purposes of the federal Computer Fraud and Abuse Act.
That ruling did not settle every question around AI agents and retail websites. It also did not mean every agent has a permanent right to shop anywhere. It did show that old computer access laws may not answer the agent-commerce question cleanly, especially when a person is using a tool to act through an account.
Amazon can still enforce site rules technically, commercially, and through contracts. AI companies can still argue that their agents are acting for users. Courts, regulators, payment networks, and merchants now have to decide where those claims meet.
What businesses should watch
For retailers, the immediate question is whether to block, allow, or negotiate with shopping agents. Blocking protects control over the customer experience and data, but it may frustrate customers who want agents to handle routine buying. Allowing agents may increase sales, but it also introduces harder questions about fraud detection, attribution, sponsored placements, returns, and support.
For AI companies, the pressure will be toward clearer agent identity and permission models. If a tool browses as if it were a normal human visitor while actually acting as a third-party agent, large platforms are likely to resist. If agents identify themselves, merchants can make participation conditional.
For payment companies and banks, the challenge is delegated authority. Agentic commerce needs records of what the customer approved, how much could be spent, which merchant was involved, and whether the agent stayed inside those limits.
For consumers, the near-term result may be uneven access. Some sites may work with preferred agents. Some may build their own agents. Others may block outside agents until standards, contracts, and liability rules mature.
Amazon’s block is an early signal
Amazon blocking Muse is not the end of AI shopping. It is an early signal that agentic commerce will not be decided by model capability alone.
An AI agent may be useful, secure, and authorized by a customer, yet still be rejected by the service it tries to enter. That is the unresolved problem. The web has spent decades asking whether a visitor is human or bot. Agentic commerce introduces a harder category: a bot authorized by a human, rejected by a merchant, and connected to a payment card.
Until the industry agrees on identity, consent, and accountability, more agents are likely to run into the same wall Muse hit at Amazon.
Get new small business insights by email
Practical ideas and useful articles to help you make better business decisions.
Tech Help Canada Staff researches, writes, and reviews practical content for business owners and professionals. Our coverage spans business, marketing, SEO, technology, and the tools and systems people use to grow and operate online. We focus on clear, useful information backed by research, hands-on experience, and editorial review. Learn more about our team and editorial standards. Need help with something? Contact Us
