NEW YORK — JPMorgan Chase and Wells Fargo have become the latest big banks to warn customers over potential liabilities risks related to giving AI agents access to their accounts, saying they need to understand the potential risks, including how responsibility for misuse of their data could shift.
As the CU Daily reported earlier here, Bank of America has also warned customers not to share their online banking usernames and passwords with artificial intelligence tools that offer to act on their behalf, while Canada’s largest bank is blocking AI agents.
Speaking during the bank’s third-quarter earnings call, CEO Jamie Dimon said JPMorgan is restricting access by outside AI agents while developing tools to protect customers, according to Banking Dive.

“You don’t want rogue agents, and there could be thousands of them,” Dimon said. “If people are using them, they should understand the risk.”
The warning comes as AI assistants increasingly promise to manage consumers’ money, potentially moving deposits between institutions to capture higher interest rates. Banking Dive reported that the launch of Meta’s Muse assistant renewed questions about whether such tools could disrupt banks’ deposit relationships.
Dimon said regulators should examine whether widespread use of autonomous agents could create new liquidity risks for banks.
“It’s a new world,” he said, adding that regulators should prepare for the possibility of problems involving bots and agents.
AI Spending and Returns
JPMorgan is testing Smart Cash, an AI-powered service designed to move customers’ funds among accounts within the bank to improve yields and manage cash flow, Chief Financial Officer Jeremy Barnum said.
Barnum said the bank’s annual AI spending was closer to $700 million than the $2 billion figure previously discussed, although he described the estimate as outdated and difficult to isolate from broader technology costs.
He also said benefits from newer generative and agentic AI tools are not yet visible in the bank’s quarterly financial results, even as internal experiments show encouraging productivity improvements.
Bank Seeks to Control AI Costs
JPMorgan’s third-quarter expenses rose 12% from a year earlier to $27.3 billion, reflecting higher compensation, staffing, brokerage, marketing and technology costs, Banking Dive reported. Barnum projected fourth-quarter expenses of about $27.5 billion.
To control AI costs, the bank is seeking flexibility to use proprietary and open-source models and route tasks to less expensive systems when appropriate.
“You don’t need the most expensive things for every query you have,” Dimon said.
Wells Fargo Also Cautions Customers
Meanwhile, Wells Fargo is also warning customers they could be responsible for mistakes made by artificial intelligence agents given access to their accounts.
The bank’s revised access agreement, effective Nov. 24, says instructions and transactions initiated by an AI agent using a customer’s credentials will be treated as authorized by that customer, unless otherwise provided by law or regulation.
Customers will be responsible for monitoring agents and for errors, unintended actions and transactions that exceed the authority they granted, according to the agreement.
Wells Fargo also delivered the warning in an Oct. 2 customer email, Payments Dive reported.
The bank told customers that sharing login credentials with an AI tool could leave them liable for the tool’s mistakes, according to the publication.
What Agreement Says
Wells Fargo’s agreement says the bank does not monitor or review agent activities. It also says the bank does not control how outside providers collect, store or share customers’ information.
The provisions include exceptions for applicable law and regulations, meaning the agreement does not establish unconditional customer liability in every circumstance.
A Wells Fargo spokesperson told USA Today the bank has taken steps to block unauthorized agents from accessing customer accounts.
Other institutions are imposing restrictions. Citigroup told USA Today it does not permit outside AI agents to log into customer accounts or transact within its digital banking platforms, while allowing limited, read-only access through approved providers.





