On Tuesday, during JPMorgan Chase’s third-quarter earnings call, CEO Jamie Dimon drew a hard line against outside AI tools acting for bank customers. “We’re not going to allow just any outside agent to come inside the company, because that’s just too risky for clients and for JPMorgan,” he said. “You don’t want rogue agents.”¹

The bank later confirmed that it blocks third-party agents that try to access accounts using customers’ banking credentials. Its argument is straightforward: an AI agent capable of logging in and taking action can also make costly mistakes, leak sensitive data or become a new route for fraud. JPMorgan says it will still support customer-authorized data sharing through open-banking providers, but not an unrestricted parade of automated agents at its doors.¹

Dimon’s intervention lands as consumer AI companies push assistants beyond drafting emails and booking trips. Meta’s Muse, launched in September, offers financial guidance through a Plaid integration, allowing users to share balances, transactions, investment holdings and mortgage information. Plaid has framed the pitch as a more natural, conversation-led way to manage money, based on information customers choose to provide.¹

The clash is not simply banks versus automation. Some firms are trying to prove that AI can operate within strict guardrails. In August, Germany’s Scalable Capital opened its platform to models including ChatGPT, Claude and Grok for portfolio analysis, savings plans and trade placement — while requiring users to approve trades and barring payments or withdrawals. Its product chief, Alexander Seipp, called the technology a way to put information, computing power and intelligence “literally in your pocket, 24/7.”¹

But JPMorgan’s scale makes its caution consequential: the bank reported $5.1 trillion in assets as of Sept. 30. And concerns about autonomous systems are no longer theoretical. A Cloud Security Alliance note described a campaign in which open-source AI agents allegedly helped breach at least 27 organizations and steal more than 600,000 card records from two of them.¹

For now, JPMorgan is betting that the fastest route to trustworthy AI in banking is to keep outside agents on the other side of the login screen. The Verge separately reported that the bank confirmed those agents cannot perform customer tasks — a blunt signal to an industry built on making software act for people.²