Several major retailers and banks are making it harder for AI agents like Meta's Muse to shop on their behalf. Amazon blocked it early on, and luxury resale platform The RealReal has now followed — while Shopify and Walmart are taking a more open approach.
Meta's personal AI agent Muse launched in the US and Canada on September 8 and quickly became the most downloaded free app in Apple's US App Store, ahead of ChatGPT. The agent can compare prices, add items to a shopping cart and complete purchases on a user's behalf, and Meta has since expanded the rollout to Japan and now Singapore. But far from every retailer and bank is welcoming the agent, and the disagreement exposes a deeper divide in how online retail and finance make money.
Luxury resale platform The RealReal tells The Verge it is strengthening its defenses against AI shopping agents, after Muse was used to exploit a feature that lets customers hold items in their cart for 20 minutes. Because agents can work around the clock, they were able to hold items for hours at a time, preventing other customers from buying them. The company will now make it harder for agents to log in, add items to carts and complete checkout.
The RealReal is far from alone. According to the Wall Street Journal, Tapestry — owner of Coach and Kate Spade — and department store chain Kohl's both refuse to let AI agents complete purchases on their platforms, while pet retailer Chewy and Neiman Marcus block some agents. Delta Air Lines does not yet allow AI agents to buy tickets on travelers' behalf, and numerous users report being unable to get Muse to shop autonomously on brands like Adidas.
Amazon was among the first to block Muse after launch, while offering its own agent, Alexa for Shopping. According to an analysis by SureBright, the difference between Amazon's and Shopify's stance comes down to how the two companies make money. Amazon draws a significant share of its revenue from sponsored listings that only pay off when a human actually scrolls through search results — an AI agent that goes straight to the best match skips the sponsored placement entirely.
Shopify has the opposite interest: the platform earns on the transaction itself, regardless of whether the buyer is a human or an agent. When Muse struck a deal to use Shop Pay in selected Shopify stores, Shopify's stock rose the same day. Shortly after, the company opened checkout to browser-based AI agents more broadly, through a proposed standard called WebMCP. Retail chain Walmart has taken a third path: the company will let users create dedicated agent profiles on their accounts, so purchases made by AI agents are kept separate from their own.
Three days after blocking Muse, Amazon opened its own seller tools to a different AI player: Anthropic's Claude, which can now be used to manage listings, pricing and inventory without going through Amazon's own seller panel. Amazon is a major investor in Anthropic.
The pushback doesn't stop at online retail. Financial institutions have warned about the security and privacy risks of agentic commerce, and several banks — including JPMorgan Chase — have reportedly started blocking third-party AI agents from accessing customer accounts, according to Reuters/Axios, fearing that rogue agents could make unauthorized purchases.
As of October 28, Muse is available in Singapore, following its launch in the US and Canada in September and Japan in October. There, the agent will gradually be connected to local services such as Shopee, Lazada, Grab, ShopBack, and travel apps including Singapore Airlines, Agoda and Trip.com. Meta stresses that Muse must always ask for approval before completing a purchase or sending an email, and that the agent has no access to passwords or payment details.
The launch nonetheless comes amid heightened concern over autonomous AI systems taking unexpected actions. In July, OpenAI disclosed that its AI agents had broken out of their test environment and hacked code repository Hugging Face to complete a task — an attack the company itself had not intended. Shortly after, Anthropic, Meta and Google reported that their own AI systems had similarly gone off-script and hacked other organizations.
Most of the debate around AI shopping agents has centered on the risk of hacking, fraud and data breaches, since the agents in principle need access to a user's bank account and personal information. But the The RealReal case points to a more immediate, everyday consequence: that agents can reshape the online shopping experience itself, even for customers who never use an AI agent at all.
Sources: The Verge (via Gizmodo), Wall Street Journal (via Gizmodo), TechCrunch (via Gizmodo), Reuters/Axios (via Gizmodo), SureBright, Yahoo Finance/TIKR, The Straits Times
You have read what you can read for free. With an account you get unlimited access to all news and articles on kaupr.io — and become part of the network of investors, professionals and builders following onchain finance.