JPMorgan Chase has put a stop sign in front of the latest AI-finance pitch: letting outside agents log in and act for customers. On Tuesday, chief executive Jamie Dimon said the bank would not admit “just any outside agent” because it was “too risky for clients and for JPMorgan,” adding: “You don’t want rogue agents.”¹

The stance arrives as companies market agents that can handle everyday tasks — and, increasingly, help users interpret balances, transactions and investments. JPMorgan’s concern is not merely theoretical. The bank says it blocks third-party agents that obtain customers’ banking credentials, while still supporting customer-authorized data sharing through open-banking providers. Its position reflects a narrow distinction: sharing information through approved rails is one thing; handing an autonomous tool the keys to an account is another.

That line hardened after Dimon’s earnings-call remarks. JPMorgan confirmed it blocks outside AI agents from performing tasks for customers, reinforcing that the policy is an operational barrier rather than a warning shot.²

Charles Schwab has now landed in the same camp. In an interview with Bloomberg, CEO Rick Wurster singled out Meta’s Muse and warned against clients simply giving it a password. “We can’t control it, the client can’t control it, something could go wrong and who’s liable?” he said. “For the protection of the client and for the protection of us, we’re not allowing it.”³

The industry’s argument is therefore converging around control, accountability and fraud exposure. AI companies are offering a different answer: Meta is developing the Poppy personal-agent protocol with Sierra, designed to show when an agent is acting for a customer and to secure consent. But for now, Wall Street’s largest gatekeepers are treating better identity signals as a future fix — not a reason to open the door today.