Why banks and Amazon are blocking AI agents like Meta's Muse

AI "agents" are meant to do chores for you: log in, fill in forms, shop and book travel. A growing list of companies is saying no. As Notebookcheck sums up, citing reports from Bloomberg, Fortune and The Verge, banks and big shops are closing the door, and one name keeps coming up: Muse, Meta's personal agent.

What is an AI agent, and why is it a problem?

A chatbot answers questions. An agent acts: it opens websites, signs in with your details and clicks buttons for you. Meta's Muse can shop, fill in forms and book trips. OpenAI and Anthropic offer similar tools in the browser, such as the ChatGPT Chrome extension. For a company on the other side, a login from an agent looks like you, but it isn't you, and nobody fully controls what it does.

Banks: no outside agents near the money

• Charles Schwab. CEO Rick Wurster told Bloomberg TV on October 16 that the broker stops dozens of attempts by major AI agents every week to reach client money, and he gets a weekly report on them. He named Muse. Outside AI models will be able to check balances and pull research through Schwab's own MCP servers (a standard way for AI tools to connect to services), but they can't move money. For everything else, Schwab points clients to its in-house assistant, Charley.
• JPMorgan. CEO Jamie Dimon said on the bank's October 13 earnings call: "We're not going to allow just any outside agent to come inside the company." A spokesperson told Fortune the bank blocks any third-party agent that tries to sign in with a customer's banking credentials. Customer-approved data sharing through open-banking providers is still allowed.

Wurster put the reason simply: "We can't control it, the client can't control it, something could go wrong and who's liable?"

Shops and travel: same story

According to The Verge's October 17 roundup, Delta is preventing agents from booking flights and Yelp only lets in agents whose makers pay for its data. Amazon said in September that Muse's continued access breaks its site policies, and users say other assistants hit the same wall there. Meta's argument is that blocking a personal agent means blocking the customer behind it, and it has signed partnerships with Walmart, Best Buy and Sephora instead.

In Europe, Amazon's German terms of use already have a section on agents: an agent must identify itself by name in every request, may not imitate human typing or solve CAPTCHAs, and can be restricted at Amazon's discretion.

What this means for you

The practical risk is not that an agent is blocked; it's what happens when it isn't. If you hand an agent your banking password and the account is misused, your bank may treat that as your negligence. Notebookcheck points out that standard German online-banking terms require customers to keep their PIN and credentials secret, with exceptions only for licensed payment and account-information services, which an AI agent is not. Dimon made the same point: if people use agents, "they should understand the risk."

Anthropic's own help page for Claude in Chrome advises against using it to manage financial accounts, prohibits stock trading, and recommends a separate browser profile without access to sensitive accounts. That's sensible advice for any agent:

1. Never give an AI agent your banking login.
2. Use the agent's manual-approval mode so you confirm each step.
3. Keep agents in a separate browser profile, away from your bank and email.

Expect more official, limited connections like Schwab's MCP servers, and fewer agents logging in as you. Meta, for its part, keeps pushing ahead with new models built for agents.