Major U.S. banks are detecting unauthorized activity from third-party AI agents operating within their customers' online accounts, according to a new Axios report.
uncertain
banks → detect → report
Some of the biggest financial institutions in the U.S. say customers have given their personal AI agents, which are unvetted by banks, access to their financial accounts.
asserted
which → say → accounts
"In a lot of cases, this is just going to be customers giving their username or password to an AI agent that the bank hasn't had a chance to vet yet, that goes against the bank's terms of service," Madison Mills, senior AI reporter for Axios, told CBS News.
asserted
Mills → go → News
"In other cases, in our reporting, we did find evidence that appeared to look like agents taking actions in customer accounts that were not authorized.
asserted
that → find → accounts
Examples of agent behavior that a user never approved include the AI calling a bank with a complaint without first consulting the account holder.
asserted
AI → approve → holder
"So, really a nightmare of AI agent activity for banks," Mills said.
asserted
Mills → say → banks
To be sure, these agents weren't going rogue or hacking customer accounts and depleting their funds, Mills said.
asserted
Mills → go → funds
They were largely following user instructions without the bank's permission to transact on behalf of a human account holder.
asserted
They → follow → holder
"In no way cyberattacks"
White House AI czar David Sacks responded to the Axios report, saying in a post on X that there's a clear distinction between personal AI agents that help manage users' bank accounts and hackers deploying AI for nefarious purposes.
asserted
that → respond → purposes
"There's a world of difference between user-authorized agents (Muse, Instinct, Grokbot, etc.) and true unauthorized/rogue agents that could constitute cyberattacks," he said.
uncertain
he → be → cyberattacks
"Most of the activity described in the Axios story appears to be the former: customers sharing credentials with agents so they can take actions on their behalf."
asserted
they → describe → behalf
Sacks added, "This may violate bank terms requiring humans to log in themselves, but these are in no way cyberattacks."
He believes banks should support the use of such user-authorized AI agents, but that if banks want to limit their access, they could grant AI agents "read-only" permission, meaning they could view a user's spending and other bank activity but couldn't take autonomous actions.
Under such circumstances, an agent could view your purchases and opine on whether you're spending too much money eating out, for example.
uncertain
you → add → example
But it couldn't make a purchase on your behalf or take any other action.
asserted
it → make → action
"Full write access (with user credentials) is a different risk category and should be approached carefully," he said.
asserted
he → approach → credentials
Mills noted that it's an industrywide problem facing dozens of financial institutions as they race to curb AI agents' access to customer accounts.
asserted
they → note → accounts
Banks taking action
Some banks said they're taking action to make their customers aware of the risks of sharing their credentials with AI.
asserted
customers → take → AI
"Customers should not give their username and passwords to anyone.
asserted
Customers → give → anyone
We support secure, customer-authorized data sharing through open banking providers to meet customers' needs while keeping their accounts safe," a Chase spokesperson told CBS News.
asserted
spokesperson → support → News
JPMorgan Chase CEO Jamie Dimon expressed concern about AI agents taking actions in customers' accounts on the company's Oct. 13 earnings conference call.
asserted
agents → express → call
"We're not going to allow just any outside agents to come in inside the company because it's just too risky for clients and for JPMorgan," he said.
asserted
he → go → JPMorgan
Charles Schwab's CEO earlier this month said the company has been blocking unauthorized agents, according to Reuters.
uncertain
company → say → Reuters
A spokesperson for Morgan Stanley said it had not observed any unauthorized AI agent activity across its platforms.
asserted
it → say → platforms
Bank of America and Wells Fargo did not immediately respond to CBS News' requests for comment on the matter.
asserted
Bank → respond → matter
The Aporia
