Banks are worried AI agents will start going on rogue shopping sprees with our credit cards.
Driving the news: The head of AI at RBC says the lender won’t allow autonomous agents to access its systems until it can ensure proper safeguards are in place, following reports that agents are causing serious security issues at U.S. banks and even taking rogue actions.
Per Axios, big banks including Morgan Stanley, Bank of America, and JPMorgan Chase have already had agents operating within customers’ accounts, and some have attempted to take actions that they were not instructed to. Some of the lenders are now restricting access as a result.
RBC’s CEO Dave McKay said this week that the lender is considering banning Meta’s Muse from accessing user information. BMO’s head honcho Darryl White issued similar warnings.
Why it matters: It doesn’t exactly inspire confidence to hear some of the world’s largest financial institutions admit that they don’t know how to host AI agents safely within customer accounts.
Even without the rogue behaviour, these agents pose a serious problem to bank security, which relies on human verification to log into accounts. Lenders say they are having a hard time differentiating fraudulent activity from authorized AI agents.
Our take: Giving AI agents the keys to our bank accounts opens up a can of worms legally. If an agent goes rogue and spends thousands unprompted — or worse, steals from a customer — who’s on the hook for the losses? The AI company, the bank, or the user?—LA
