Amazon blocks Meta’s Muse agent, opening a fight over who owns the customer when AI does the shopping

Amazon has blocked Meta’s Muse agent from its store. GeekWire first reported it, and Yahoo Finance picked it up. When Muse tried to complete a purchase, Amazon began showing a warning that continued access by an unauthorized AI agent violates its terms of use. Amazon says it had already asked Meta to keep its site out of Muse’s reach. Its argument is that third-party apps that buy on a customer’s behalf should operate openly and respect each merchant’s choice to participate or not. It compares the situation to delivery apps and online travel agencies, which work with the consent of the merchants and airlines they transact with.

Meta says Muse cannot see users’ passwords or payment methods, and that shared credentials are kept in secure storage. According to GeekWire, the agent runs on a dedicated virtual machine in Meta’s cloud. It generates single-use card numbers through Link by Stripe, and a monitoring system called Sentinel limits what actions it can take online. The two companies are also partners. Amazon products have been purchasable inside Facebook and Instagram since 2023, and in April Meta agreed to run AI workloads on Amazon’s cloud.

Amazon is not alone. According to MaisTecnologia, several airlines are blocking outside agents during flight search or booking. JPMorgan Chase, Bank of America and Citigroup are restricting automated bots from customer accounts, citing security and fraud prevention. Amazon has also done this before. GeekWire notes that it took legal action against Perplexity over its Comet browser and has targeted shopping agents from Google and OpenAI. The infrastructure layer is shifting as well. Cloudflare Radar measured bots at 60.6% of requests for HTML content in the seven days before August 10, 2026, versus 39.4% human traffic, according to Search Engine Journal. Since September 15, TechCrunch reports, Cloudflare’s default settings block mixed-use crawlers (those combining search, training and agent tasks) on any page that shows ads, unless the site owner opts in.

The official reason is security, but the real conflict is economic. An agent that compares prices and goes straight to checkout skips sponsored listings, recommendations and the page design built to grow the basket. Amazon’s advertising business brought in more than $68 billion last year, and it depends on a person scrolling. In banking, an assistant that automatically moves idle cash into higher-yield products raises what a bank pays for deposits. Search engines went through the same tension. The difference now is that the software doesn’t just read the page. It acts.

Our reading: this is the agent-era version of the fight over who controls the plumbing, which we flagged when the battle moved from models to distribution and agent standards. In the short term, expect friction: CAPTCHAs, dropped sessions, legal threats and walled gardens, with legitimate users sometimes caught in the crossfire. Neither side has a clean case. Amazon defends merchant consent, but also an ad model that benefits from human inefficiency. Meta defends user choice, while building its own position as gatekeeper. The likely way out is the one Amazon itself points to: opt-in, authenticated agent protocols, similar to how travel agencies and delivery apps work. The long-term prize is real, because agents that compare prices honestly lower costs for consumers, and lower costs fit our idea of abundance. But first the web needs a way to tell an authorized agent from a malicious bot, and a business model that doesn’t require a human looking at the page.