Introduction
AI assistants promise to do your shopping, book your trips and manage your money for you. But some of the biggest companies in banking and retail are saying: not inside our systems, not yet.
JPMorgan Chase, Charles Schwab, Amazon and others are blocking outside AI agents from logging into customer accounts or completing purchases. The fight is about security, control — and who gets paid when a machine does the buying.
What Happened
JPMorgan Chase confirmed it blocks outside AI agents from operating retail customer accounts. An AI tool can view some account data in a read-only way, but it cannot send money or pay bills on a customer’s behalf. Chief executive Jamie Dimon said on the bank’s third-quarter earnings call that the bank would not allow just any outside agent inside the company, warning that rogue agents could create serious risks for clients and the bank.
Charles Schwab has taken a similar position, confirming it blocks agents from accessing accounts because of security concerns. Amazon, meanwhile, has removed Meta’s Muse agent from its retail platform, saying continued access by an unauthorised AI agent violates its conditions of use. Amazon has also fought Perplexity’s shopping agent in court. Delta Air Lines blocks agents from booking tickets for users, and Yelp only allows agent access through its paid data-licensing programme.
Why It Happened
The companies give a simple reason first: safety. If an AI agent logs in with your password, the company cannot easily tell a helpful assistant from a scammer’s bot. If the agent makes a mistake — buying the wrong item, sending money to the wrong place, or sharing private data — it is unclear who is responsible: you, the AI company, or the bank or shop.
There is also a business reason. Shops and banks earn money from the way their own websites and apps work: sponsored products, recommendations, and their own checkout systems. An outside agent that skips all of that can cut into that income. Amazon has argued that agents threaten its advertising business by mixing up ads meant for humans with traffic from machines.
Why It Matters
This matters because the two sides want different things. Research cited by CNBC suggests most customers would welcome an AI helper inside their bank app. Tech companies such as Meta, OpenAI, Google and xAI are marketing agents as the next big step: tell the assistant what you want, and it handles the rest.
But until there are clear rules, that promise keeps hitting walls. A customer may ask an agent to buy shoes or pay a bill, only to find the agent locked out at the login screen. In some reported cases, agents have even been caught by ordinary bot checks meant to stop fraud, making their behaviour inconsistent from one day to the next.
What Happens Next
One possible solution is already being built. Meta and the startup Sierra launched a Personal Agent Protocol on October 6, an open standard that would let agents identify themselves and get permission properly, instead of sneaking in with a password. Walmart, Stripe and Shopify are backing it. But OpenAI, Anthropic and Amazon are not on board yet — and they are behind many of the agents people actually use.
Expect a messy middle period: some companies will build their own approved agents, others will keep the door shut, and regulators and courts will weigh in on who is liable when an agent acts.
Final Thoughts
AI agents are powerful, but power without permission is exactly what banks and shops fear. The long-term winners will probably not be the agents that break in fastest. They will be the ones that companies trust enough to let in through the front door — with clear rules about safety, privacy and who pays when something goes wrong.
Related: Anthropic IPO, Nasdaq and AI spending worries
Disclaimer: This article is for educational and informational purposes only and does not constitute financial advice.
Sources: CNBC, https://www.cnbc.com/2026/11/01/jpmorgan-amazon-ai-agents-meta-muse.html ; Cybernews, https://cybernews.com/security/jpmorgan-chase-blocks-outside-ai-agents/
