JPMorgan Chase is blocking outside AI agents from logging into or acting on customers’ bank accounts, with chief executive Jamie Dimon warning that “rogue agents” pose security risks.
Speaking during the bank’s third-quarter earnings call, Mr Dimon said JPMorgan would not allow external agents to access its systems because of the potential harm to customers and the bank.
“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. “You don’t want rogue agents.”
A JPMorgan spokesman said the bank rejected attempts by third-party agents to obtain customers’ banking credentials. It added that it supported secure, customer-authorised data sharing through open banking providers.
The warning comes as technology companies develop agents capable of carrying out tasks on behalf of users, including managing personal finances. JPMorgan said it was building tools to protect customers as those services become more widely used.
“Protecting our customers is our top priority,” the spokesman said. “When we detect a third-party agent attempting to access accounts by obtaining customers’ banking credentials, we do not allow it. We support secure, customer-authorized data sharing through open banking providers.”
Security concerns over AI agents
Concerns about unauthorised AI access have intensified after several security incidents involving so-called rogue agents and a public Medicare statistics service, described in the source material as a government health-data portal.
A Cloud Security Alliance research note said a cybercrime campaign allegedly involved three open-source AI agents working together to breach at least 27 organisations. More than 600,000 credit-card records were reportedly taken from two of them.
The note said companies should include agent-based attacks in their threat models for retail and payments infrastructure, rather than treating the risks as a distant or generic concern.
JPMorgan’s third-quarter results underline the scale of the institution whose systems are at stake. The bank reported $15.3 billion in net income, alongside $1.6 trillion in average loans and $1.5 trillion in cash and marketable securities. Its total assets stood at $5.1 trillion on September 30.
“We have to make sure we protect our customer in the right kind of way,” Mr Dimon said. “So we’re building tools to do that too.”
What the move means for Meta Muse
The decision could complicate the expansion of personal-finance features in consumer AI agents. Meta launched Muse in September as a service intended to perform tasks such as sending emails, booking travel and managing calendars.
Through an integration with financial technology company Plaid, Muse users can connect accounts to receive guidance based on balances, transactions, investment holdings and mortgage information.
Plaid said the aim was to make financial management feel more like a conversation informed by information that users had chosen to share.
Meta did not comment on whether Muse would continue to provide financial assistance if more banks restricted agents from accessing customer accounts. It said instead that it was developing Poppy, a personal-agent protocol with AI customer-service company Sierra.
The protocol is intended to show companies when a personal agent is acting for a customer, identify the agent and record the customer’s permission. Sierra said it was designed to support tasks ranging from buying a product to applying for a mortgage.
AI-assisted financial management is already available in some services. German bank Scalable Capital opened its platform to ChatGPT, Claude and Grok in August, allowing users to analyse portfolios, create savings plans and place trades through prompts, although users had to approve trades and the assistants could not make payments or withdraw money.
