Analysis
The largest U.S. bank is drawing a line between AI tools it builds and controls and outside agents that customers might send to act on their accounts. That matters as consumer AI assistants start to promise to pay bills, move money and shop, because a bank that refuses those agents, or makes customers carry the risk, shapes how far they can go. Dimon's comments also add to a run of bank and regulator warnings that the newest AI models make cyberattacks easier.
On the earnings call
An analyst asked JPMorgan's executives on the October 13 third-quarter earnings call about the difficulty of using AI agents inside a bank as so-called rogue agents spread [1][2]. Chief financial officer Jeremy Barnum noted that several personal AI agents had been released during the quarter [1].
Dimon said the bank uses agents internally and will keep doing so, and that it is "not against agents or for agents" [1][2]. But he said "you do not want rogue agents, and there could be thousands of them," and that JPMorgan is "not gonna allow just any agent to come, outside agent to come inside the company" [1]. He also raised who bears the losses when an agent goes wrong, describing it as a liability transfer, and said the bank has to protect its customers [1][2].
Later, another analyst asked whether AI agents could make deposits move faster in a crisis. Barnum said it is easy to imagine cash becoming even easier to move under some kinds of stress, and Dimon added: "with all these bots and agents and stuff like that, could someone get hurt and cause a problem?" [1]. Answering a separate question, Barnum said some services that let agents move customers' money rely on customers sharing their bank login credentials [1].
The Bloomberg interview
In a Bloomberg TV interview on October 6, Dimon said AI-related cyber threats "went up 10-fold after Mythos," referring to Anthropic's most advanced model, which the company has shared with a limited group of organizations rather than releasing publicly [3][4][5]. "AI created vulnerabilities that we didn't know about, and we always worried about cyber before these things," he said [4].
He said he was "not going to get hysterical" over whether the threat is existential, and that the bank is "rolling up our sleeves and going to work to fix it" [4][5].
What remains uncertain
The call quotes come from third-party transcripts, which differ slightly in wording; JPMorgan's own webcast was not checked, so only short passages consistent across transcripts are quoted. Dimon did not say what specific controls JPMorgan uses on outside agents or how widely it blocks them. The Bloomberg interview was read as quoted by other outlets, not from Bloomberg directly, and the tenfold figure is Dimon's own estimate, not a published measure.
Sources
- The Motley Fool: JPMorgan Chase (JPM) Q3 2026 earnings call transcript, October 13, 2026 ↗Source
- Investing.com: Earnings call transcript, JPMorgan Chase Q3 2026, October 13, 2026 ↗Source
- Bloomberg: JPMorgan CEO Dimon Says Anthropic's Mythos Pushed Cyber Risk Up 10-Fold, October 6, 2026 ↗Source
- PYMNTS: JPMorgan CEO Warns AI Risks Jumped Tenfold After Mythos, October 6, 2026 ↗Source
- The Motley Fool: JPMorgan Chase CEO Jamie Dimon Just Issued a Blunt Warning About Artificial Intelligence, October 8, 2026 ↗Source