Two of the biggest names in American finance said last week that outside AI agents are not welcome inside customer accounts. Jamie Dimon made the point first, on JPMorgan Chase’s earnings call on Tuesday, October 13. Charles Schwab chief executive Rick Wurster followed two days later and repeated it to Bloomberg Television. The timing makes sense, because Meta’s Muse agent, which can shop, book and negotiate for users, arrived about a month earlier. For now, banks blocking AI agents is a stated policy more than a visible crackdown. Still, it shows where the line is likely to sit when you try to hand your money to a bot.
This article covers what the two banks said, what they will allow instead, why money is the sticking point, and what to do if you want an AI agent to help with your finances.
What Schwab and JPMorgan said about AI agents
Start with Dimon. According to The Next Web’s report on the two banks, he told analysts that JPMorgan won’t let an outside agent act inside its systems, and he warned about “rogue agents.” The bank already lets some third-party tools view account data, but they cannot send money or pay bills.
Schwab was more specific. On the Thursday, October 15 earnings call, Wurster said an agent that logs in with a client’s own credentials, or drives the website the way a person would, isn’t safe. “We are not going to allow that,” he said, as American Banker reported. The paper added that Schwab gets a notice each time it blocks one, and that such notices are rare because most clients don’t use AI that way yet.
The next day, Wurster told Bloomberg that the blocked attempts run in the tens or low twenties a week. He read that as proof that customers aren’t ready to hand AI the keys. His bigger worry was liability. If something goes wrong, who pays? Congress is still debating liability for AI agents, so nobody has a settled answer.
What the banks will allow instead
Neither bank is saying no to AI. They’re saying no to a particular way of connecting it. Schwab says it will offer MCP servers, which are a standard way for an AI model to connect to a company’s systems under that company’s rules. Through them, an outside tool can check balances and do research, but it cannot move money.
Schwab also wants clients to use its own assistant, Charley, which it began rolling out this month. Charley runs on outside language models plus Schwab’s data. At first it handles a short list of actions, such as changing an address or naming a trusted contact. Wurster expects it to handle the full range of schwab.com tasks, by text or voice, over time. Clients will eventually be able to trade through other AI tools, he said, but not move money, which he called too risky for clients and for the firm. That split, with banks blocking AI agents from acting while letting them look, is the pattern to remember.
There’s a wider push to set the rules for software that shops on your behalf. We covered one piece of it in our look at Meta and Sierra’s effort to decide how shopping bots get through the front door.
Banks blocking AI agents: why money is the line
Passwords are the first problem. When you give an agent your login, the bank sees a normal sign-in. It can’t tell you from the software, and it can’t tell a helpful request from a hijacked one. The Next Web adds that US and European regulators are asking banks about their own use of agents and the cyber risks that come with them. It also notes that MCP connections have had security flaws of their own, including one that Google and JPMorgan, among others, had to fix.
Credentials are also what criminals go after. Our report on the suspected AI-driven hacks of South Korean banks argued that cheap credential attacks, not exotic AI, were the real story.
The second reason is less comfortable, because banks blocking AI agents also protects a revenue stream. Schwab holds about $493 billion in cash-sweep accounts, according to American Banker, and those balances helped produce about $3.48 billion of net interest revenue in the third quarter. Some analysts worry that agents could shift idle cash to higher-paying options. Wurster disputes that and says people who use AI hold more cash, because they trade more. One consultant quoted in the same report said the data worries are real, but that Schwab also has a business reason to keep customers inside its own tools. Both things can be true. A bank can have a sound security case and a financial interest in the same policy.
What banks blocking AI agents means for you
If you’ve been tempted to let an AI agent run your accounts, the direction of travel is clear, and a few habits will keep you out of trouble.
First, don’t give your banking or brokerage password to any agent or browser add-on that offers to log in for you. Read-only connections, where the tool can see balances but can’t act, are a far safer start, and more banks are likely to offer them. Second, check what you’re approving. “View my accounts” and “move my money” are very different permissions, and the second should be rare. Third, turn on transfer alerts, so a surprise payment shows up on your phone within minutes.
Also check your account agreement. Many banks tell customers not to share login details. How a dispute would be handled if an agent made a costly mistake is exactly the question Wurster raised, and no one has answered it yet. Businesses face the same problem, which we looked at in our piece on the new AI exclusions in liability insurance.
This is general information, not financial or legal advice.
What to watch next
Three things are worth tracking. The first is whether other large banks join the group of banks blocking AI agents in public statements, since only two have so far. The second is how fast Charley and similar in-house assistants expand beyond simple tasks. The third is Congress, because a clear liability rule would change how willing banks are to open the door. If the rule puts the risk on the bank, expect more blocking. If it puts the risk on the customer who authorized the agent, expect more consumer warnings.
Editor’s take
Banks blocking AI agents is both protective and self-interested, and you don’t have to pick one explanation. For ordinary customers the practical answer is the same either way: keep AI agents on a read-only leash until the liability question is settled. Use them to summarize, budget and research, and keep the final click on any transfer for yourself. When a bank does offer a sanctioned connection, that’s the one worth trying first.
