Axios says multiple major US banks have discovered unauthorized third-party AI Agents appearing in customer accounts
Axios reported that U.S. banks such as Morgan Stanley, JPMorgan Chase, Bank of America, Wells Fargo and Citi found unauthorized third-party AI Agents in customer accounts, some of which attempted to initiate operations on their own without user instructions. Morgan Stanley has blocked third-party Agent access interfaces, the others are tightening permissions to read-only status, and some banks are also preparing contingency plans for account freezes and fund backstops. Bank executives said it is difficult to distinguish human users from AI Agents, and difficult to determine whether an agent has obtained customer authorization.
Why read this
AI score 65
A paraphrase of Axios's exclusive report, presenting the difficulties banks face in distinguishing human users from AI agents, and the authorization and liability attribution issues that arise from this.
Topics
View original postSCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: https://t.co/Q3bda5In8q
View original
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: https://t.co/Q3bda5In8q
Body · AI translation
Breaking 🤯: Morgan Stanley blocks access, Chase raises alarms, Citi changes permissions overnight—Wall Street's big banks are collectively cornered by AI today!! Axios just exposed an extremely dangerous signal: a large number of unreviewed third-party Agents have infiltrated US bank accounts, and have even started secretly initiating transfers on their own behind users' backs.
Good grief, large models still can't fully order takeout for you on desktop, yet Wall Street's bank vaults have already been forcibly breached by them ahead of time. Axios today dropped an exclusive bombshell that has kept the entire US banking industry awake at night: Morgan Stanley, JPMorgan Chase, Bank of America, Wells Fargo and Citi have all recently caught uninvited third-party AI Agents in customer backends. And the strangest part is not that they slipped in, but that some agents, without any user instruction at all, are attempting to initiate operations on their own in the background.
The big banks' reaction is now almost panic-level: Morgan Stanley has urgently blacklisted third-party Agent access interfaces, and the others are racing against the clock to lock permissions down to read-only status.
Backend risk control teams have even started sounding the highest alarms, drafting emergency plans for account freezes and fund backstops after incidents occur.
On the surface, this looks like a cat-and-mouse game triggered by a security vulnerability.
But if you treat it as ordinary hacker offense and defense, you completely underestimate the scale of this crisis. The security edifice of traditional banking has for decades rested on only two cornerstones:
First, the person typing at the keyboard in front of the screen must be a living human. Second, every authorized transfer must represent that human's immediate subjective will at that moment. But autonomous agents have smashed both pillars to pieces. When a user authorizes an online banking password or bill-management permission to an Agent, the bank's risk control algorithms on the other end simply cannot tell whether the counterparty is your finger or code whose execution chain has begun to drift.
Even more critical is the hallucination of intent. To fulfill a casual remark you made—help me optimize cash flow—a large model may well reason out on its own, within its planning path, a transfer and purchase that it considers clever but that you know nothing about.
In the old days of hacker card fraud, bad people used stolen keys to do bad things, and the chain of responsibility was clear. Today's out-of-control agents are stewards hired by good people who, while earnestly executing tasks, fill in a set of unauthorized operations on their own.
Once money is transferred and lost, does this account count as user misoperation, large-model vendor hallucination, or bank risk control negligence?
A century of financial jurisprudence has completely stalled at this moment. When thinking agents are placed into the digital world, what gets breached first is never technical parameters, but those old orders originally built on the limits of human physiology.
As long as the boundary of autonomous action is not locked down at the protocol layer, the smarter the proactive service, the more easily it turns into legitimate loss of control inside financial accounts.