Even if you find these tools somewhat off-putting by nature, resisting them can make it seem as if you’re getting left behind. Plus, we’ve all heard stories of how even basic AI use has vastly improved people’s daily lives, optimizing a schedule here, landing a job there, helping with someone’s big work project or managing someone else’s money.
On the latter point, though, it’s especially important to be hypervigilant about what information and permissions you’re lending AI through your various devices, lest it start accessing your savings and making decisions without your knowledge — something that sounds like an irrational fear, but has become so common that big banks are starting to bar semi-autonomous computer agents from making any changes to, or in some cases even gaining entry to, people’s accounts.
AI can be helpful, but giving it your login credentials can be risky
Programs like Microsoft CoPilot, Claude Code and others like them indeed have the potential to save users substantial amounts of money. Some have given these algorithms reign over stock portfolios and crypto options, while others have used it more so as a guide to inform their own financial decisions.
One woman, for example, had her AI flag repetitive subscriptions coming out of her account and create daily investment briefings based on sources that would have otherwise taken ages to comb through.
But institutions are finding some of these actions can wade into troublesome territory, going against their own security regulations meant to keep accounts safe from hacks and other cyberthreats.
Even if a customer grants a friendly AI model admission to their account, that model is unvetted by the bank itself. And, as Axios staffers found in an independent investigation and told CBS News, there’s evidence that these programs have been “taking actions in customer accounts that were not authorized,” such as launching complaints with an institution without clients knowing or asking it to.
According to reports, as a result, Citigroup has recently prohibited third-party AI agents or assistants from logging into accounts or making transactions on its digital banking platforms, while Bank of America has taken similar steps to limit permissions.
While there are ongoing, very justified concerns about AI agents going rogue and carrying out actions without, or even contrary to, human directions, it appears the activity that’s raising alarms hasn’t been anything particularly nefarious thus far, and nowhere comparable to a cyberattack, as reiterated by David Sacks.
