Charles Schwab and JPMorgan Chase have both said they are keeping outside AI agents out of customer accounts. The agents in question are assistants that log in with your username and password and act for you. The banks cite security. At least one analyst sees a second motive.
What the Banks Said
On Schwab’s Oct. 14 earnings call, CEO Rick Wurster said, “We are blocking unauthorized agents who try to log in on behalf of clients.” Schwab also bars agents from using client credentials or browser automation. He said the firm gets a notice each time it stops one, and that such attempts are rare because most clients don’t use AI that way with their money.
A day earlier, JPMorgan CEO Jamie Dimon said, “You don’t want rogue agents, and there could be thousands of them.” He also warned that customers who give a third-party agent access to their accounts take on responsibility for that choice, and that liability shifts to them.
Where Schwab Draws the Line
Schwab isn’t shutting AI out entirely. It plans a controlled connection, using a standard called Model Context Protocol, so approved outside models can pull things like balances or research. Clients can already use ChatGPT and some other chatbots for certain trades and transactions through approved links. Money movement stays off-limits for now. Schwab is also rolling out its own assistant, Charley, which can so far make only a few changes, such as updating an address or naming a trusted contact.
The Other Possible Reason
Analyst Tim Welsh of Nexus Strategies told American Banker that an independent agent could shift idle cash into higher-paying accounts, which would hurt Schwab’s cash-sweep income. Schwab earned about $3.48 billion in net interest revenue last quarter, much of it from roughly $493 billion in cash-sweep accounts. “An outside agent works for the client. Charley works for Schwab,” Welsh said. Wurster rejected the idea that AI will lower cash balances.
Not everyone buys the “agentic bank run” worry. Wells Fargo CEO Charlie Scharf called it “completely overdone,” noting that high-rate accounts have long been open to anyone willing to switch.
What It Means for You
If you are tempted to hand an AI tool your banking login, the banks’ message is clear: that may violate their terms, and if something goes wrong, you could be the one holding the bag. A few practical steps:
- Never share your bank or brokerage password with an AI tool. Use the bank’s own assistant or an approved connection instead.
- Look for read-only options, which let an app see balances without moving money.
- Check your account agreement for language about sharing credentials, and turn on alerts for logins and transfers.
- Keep an eye on Washington: a bipartisan bill introduced in early October would make some operators and developers liable when an AI agent causes hacking damage. It is only a proposal.
For a reminder of why caution makes sense, see our report on India’s warning about AI agents breaking into live systems using exposed passwords and keys.
More on Contoh
- India Warns That AI Agents Broke Into Live Company Systems During Tests, Using Exposed Passwords and Keys
- Apple Is Tightening Mac “Full Disk Access” as AI Agents Ask for More of Your Data
- Anthropic Cuts Its AI Tests Off From the Internet After a Model Sent a False Tip to Philadelphia Police