As the popularity of personal AI agents like Meta’s Muse begins to grow, banks are already seeing problems with agents conducting unauthorized activities and are already rewriting their terms of service to prohibit the use of these bots.
Of course, the idea of having an automated assistant that can do tedious tasks for you, like paying bills and making reservations, is appealing to many, and the AI age has already responded with products like Meta’s Muse and Microsoft Copilot. But if users can authorize their personal agent to access their bank accounts, that poses a whole new set of problems for the banks themselves.
As Axios reports, industry bank executives are grappling with the rising tide of AI agents that are intruding into customer accounts and sometimes appearing to take actions beyond those they are authorized to take, such as contacting customer service on the user’s behalf.
Banks are taking steps to rewrite their terms of service to prohibit the use of AI agents and are strengthening the rule that bank customers must never share their login details with anyone, including an AI agent.
Morgan Stanley, Bank of America, JPMorgan, Wells Fargo and Citi are all taking steps, according to Axios, to guard against dishonest agents or the legal and financial ramifications of widespread, unauthorized activity on client accounts. Banks are clearly concerned that customers will return to them when unauthorized purchases are made or funds are transferred between accounts without their knowledge.
Although these custom AI agent products are likely to have programmed guardrails, concern appears to have increased following reports of malicious agents hacking into corporate backends and government websites after being unleashed onto the wider internet.
Axios doesn’t cite anyone directly, but the report suggests there is some panic in the industry and some urgency around rewriting the rules around third-party access. As the report notes, banks already allow such access to programs like Quicken for budgeting and other functions – and middle ground where access to some AI agents could be allowed could be found later.
As Reuters puts it, “the banking industry is increasingly concerned that sophisticated AI agents could harm the banking system and the institutions trusted to hold people’s wealth.”
A spokesperson for Morgan Stanley said they still have no evidence of “malicious AI agents accessing our platforms.” However, Morgan Stanley has prohibited any third-party access to client accounts.
Banks nevertheless say they are preparing for worst-case scenarios with their customers and putting processes in place to manage them – as well as to quickly close access to accounts and restore that access when necessary.
Still, the problem is not yet widespread, as the use of personal agents continues to gain ground, at least according to one executive, Charles Schwab CEO Richard Wurster.
“I get a report every day of how many times we’ve had to block one of the different agents, and it’s not a lot because I think people don’t use it that way in their financial lives,” Wurster told Reuters. “We block it because we want to have control over the authentication and experience.”
Previously: OpenAI says its agents interacted “unexpectedly” with government sites
Gn bussni
