JPMorgan Chase is restricting external artificial intelligence agents from accessing its systems, with chief executive Jamie Dimon warning that the technology presents heightened security risks and could leave customers liable for unwanted transactions.
Speaking during an earnings call on Tuesday, Dimon said the bank would not allow unrestricted access by outside AI agents, which can perform tasks on behalf of users rather than simply answer questions. He said the risks were amplified by the possibility of thousands of autonomous agents operating simultaneously.
“The typical security concerns around any new adaptation of technology are heightened” with AI agents, Dimon said, adding that customers using them should understand the risks. He also warned of a potential “liability transfer” away from banks and the possibility that customers’ data could be misused.
The bank is developing its own AI agent tools for clients, Dimon said, but intends to maintain strict controls over third-party services in the meantime. “We’re not going to allow just any outside agent to come in inside the company because that’s just too risky for clients and for JPMorgan,” he said.
Jeremy Barnum, JPMorgan’s chief financial officer, said the bank was already blocking external agents wherever possible. He raised particular concerns about services that require customers to share their banking credentials or use screen-scraping technology to retrieve account information.
“It’s not secure,” Barnum said, describing credential sharing as an approach the bank had long opposed. Customers who provide their login details to an AI agent could become responsible for unwanted actions taken through their accounts, he said. “It’s not an ideal state of things.”
JPMorgan has also tightened controls on AI tools used internally by its employees. The bank introduced additional security measures for some engineers using Anthropic’s Claude, restricting the system’s access to internal systems, according to an earlier report by Business Insider.
The bank’s caution comes as the growing use of personal AI agents prompts concerns about their potential effects on financial institutions. Unlike conventional chatbots, these tools are designed to carry out tasks autonomously, including actions that may involve managing users’ accounts. Their popularity has increased following the introduction of Meta’s personal AI assistant, Muse.
One concern raised by Apollo chief economist Torsten Sløk in late September was that AI agents could trigger a shift of deposits from low-yield accounts into higher-yield savings products, potentially contributing to a bank run. The prospect has added to concerns about competition for deposits as customers gain new ways to compare financial products and move money.
JPMorgan executives also acknowledged a market reaction to recent AI developments. Bank shares, which had already declined following turmoil in bond markets, fell further after Meta launched Muse on 22 September. Barnum said the bank had observed “a little bit of a response in bank stocks around those releases”.
He characterised the potential threat as an extension of existing pressures from customers seeking higher returns on their deposits and competition from financial technology companies. Such trends were not new, he said, but the bank was not treating them complacently.
JPMorgan is prioritising its share of customers’ primary banking relationships rather than attempting to attract depositors solely through higher interest rates, Barnum said. The bank’s standard Chase Savings account pays an annual percentage yield of 0.01%, compared with rates of around 4% offered elsewhere in the market, although JPMorgan provides higher-yield cash products to private wealth clients.
Barnum said the bank sought to maintain parity with financial technology competitors while relying partly on its traditional branch network to defend its position. “Our perspective is that, on the one hand, it’s nothing new, but that doesn’t mean that we’re complacent about it,” he said.
