JP Morgan CEO Jamie Dimon and Meta CEO Mark Zuckerberg.
Getty Images
JP Morgan Chase Chief Executive Jamie Dimon said Tuesday that the company will not allow outside artificial intelligence agents to access customer accounts and operate on their behalf, citing risks to the company as a technology company and its customers, including: Meta Release autonomous assistant.
“We’re not going to allow any outside agents into the company because it’s too risky for our clients and for JPMorgan,” Dimon told analysts on the bank’s third-quarter earnings call.
“We don’t want rogue agents. Their number could be in the thousands,” he added, explaining the potential threat to customers. In July, an OpenAI model hacked the servers of AI company Hugging Face.
Mr. Dimon, the longtime CEO of the nation’s largest bank by assets and the de facto spokesperson for the U.S. financial sector, is reminiscent of previous battles with fintech companies over access to customer data. JPMorgan has moved to seek access from data intermediaries including Plaid, but the dispute also centered on the bank’s security concerns and the tech companies’ access demands.
A JPMorgan spokesperson confirmed to CNBC that the bank is currently blocking external AI agents from working on behalf of individual customers.
“Protecting our customers is our top priority. We do not allow third parties to obtain our customers’ banking credentials and access their accounts by logging in on their behalf because it is not secure,” the spokesperson said.
“We are committed to supporting customer-approved secure data sharing and securely enabling agent connectivity through open banking providers,” she added.
This means that if Plaid or another connector has the user’s credentials, the AI agent can read-only access some customer data, such as customer balances, from the Chase account. However, you cannot send money or pay bills. This is a beneficial action for the agent.
The spokesperson added that it is unclear when the bank will introduce safeguards to allow agents to interact directly with customer accounts.
“Transfer of responsibility”
Dimon said JPMorgan uses AI agents internally and has plans to deploy them in customer-facing applications. He said the bank’s approach is not one for or against technology, but one that ensures it serves customers securely.
Chief Financial Officer Jeremy Burnham said on Tuesday that in recent weeks, JPMorgan began targeted pilots of its AI tool called Smart Cash, which aims to predict users’ cash flows and move funds to maximize yield. It appears to be a defense against external agents.
Dimon also appears to be drawing the line on liability when a customer allows an external AI agent to act on their behalf.
“If people are using them, they need to understand the risks,” Dimon said. “They should understand that there is a transfer of responsibility.”
Personal agents have gained mainstream popularity in recent months as more autonomy is integrated into AI models, such as the ability to cancel subscriptions or use credit cards on behalf of users, for example.
Instinct, a digital concierge that relies primarily on text messaging rather than an app, launched on an invite-only basis earlier this year. The startup offering the service, San Francisco-based Spear Street Technology, was founded last year and reached a valuation of $10 billion last month.
Bank stocks took a hit in September. meta Muse Agent topped the app download charts shortly after its announcement, sparking concerns that consumers would use agents to move money from low-yield accounts to higher-yield accounts. A few weeks later, OpenAI released its own agent, Dots.
Torsten Slok, chief economist at Apollo, and others pointed to the risks that agents pose to cheap bank deposits. However, that risk would be limited if agents were not allowed to freely manipulate customer accounts.
