During JPMorgan Chase's Q3 2026 earnings call on October 13, 2026, CEO Jamie Dimon confirmed JPMorgan deploys AI agents internally and intends to expand that use for customers — while naming rogue agents as a distinct operational concern that the firm is actively building defenses against.
"We are not gonna allow just any outside agent to come inside the company," Dimon said. He described the risk as significant given the potential scale.
"You do not want rogue agents, and there could be thousands of them," Dimon said. He added that regulators should be considering whether the proliferation of bots and agents could create systemic harm.
CFO Jeremy Barnum pushed back on figures that have circulated in media coverage of JPMorgan's AI spending. He said the actual AI expense number is "significantly less" than what has been reported, referencing a figure of approximately $700 million annually.
He also noted the earlier benefit quantification that accompanied those expense figures was accurate and included established machine learning applications such as fraud prevention.
Barnum said he is "very reluctant to spend a lot of energy quantifying" AI benefits and that current gains are not yet visible in quarterly earnings. He pointed to early productivity evidence in software engineering, including a test involving legacy code rewrites, and noted that some productivity gains are being offset by token costs.
On infrastructure strategy, Barnum described an approach built around avoiding lock-in: the firm uses both proprietary and open-weight models and applies routing logic to direct queries to the most cost-effective suitable model.
"Model agnosticism, harness agnosticism, and ability to run both proprietary models" was how he framed the approach.
Bank of America analyst Ebrahim Poonawala raised the question of whether AI-powered personal agents could accelerate cash movement out of JPMorgan accounts.
Barnum framed the risk as the intersection of two existing dynamics, yield-seeking behavior and fintech competition, rather than a structurally new threat. He said the firm is not complacent and noted that a "smart cash" product went live in recent weeks as a pilot response.
On the question of customer protection, Barnum said customers who share credentials with an external agent bear the risk if that agent acts against their interests.
Barnum also addressed AI concentration risk in JPMorgan’s lending book, describing hyperscaler-related exposure as “quite modest” relative to total wholesale exposure.
He noted the AI theme has diffused broadly enough to reach working capital revolvers and equity markets, and said the firm runs stress scenarios for a sharp reversal in AI sentiment — a sign that even JPMorgan’s bullish posture on AI comes with contingency planning built in.