Skip to content
NewsIncident

Report: unvetted AI agents still reach some US bank accounts after October's lockdown

· by Pondero Newsdesk

The short version

Axios reported, in a story picked up by Reuters, that unvetted third-party AI agents are still getting into some US bank customer accounts despite a mid-October lockdown by five major banks.

Report: unvetted AI agents still reach some US bank accounts after October's lockdown

Five of the biggest US banks spent mid-October blocking outside AI agents from logging into customer accounts. Axios reported on October 22, in a story picked up by Reuters, that the blocks have not closed the gap: unvetted agents are still reaching some accounts.

What

JPMorgan Chase, Bank of America, Citi, Charles Schwab, and Wells Fargo each moved in the same week to stop third-party AI agents, including Meta's Muse and Instinct, from signing in with a customer's saved credentials and acting on their accounts. The trigger was a run of incidents involving leaked data and credentials tied to agent logins. JPMorgan Chase CEO Jamie Dimon told analysts on an October 13 earnings call that the bank would not let "rogue agents" inside, warning there could be thousands of them, per PYMNTS' report on the call. Schwab CEO Rick Wurster said his brokerage was already catching large unauthorized agents, Muse among them, attempting logins "in the tens and 20s" every week, per that same PYMNTS account of a Bloomberg report. Axios's follow-up, picked up by Reuters, reported that despite those blocks going into effect, unvetted agents are still getting into some bank customer accounts. Axios's full article sits behind a bot-detection wall that blocked Pondero's own attempt to verify it against the source text, so this brief does not repeat any bank-by-bank or agent-by-agent detail beyond what ran in the Axios headline and what Reuters confirmed in its pickup.

A five-bank block still has a hole in it

A lockdown announced by five banks in one week was never going to be airtight on day one: Schwab's own numbers show large known agents attempting logins dozens of times weekly even before the blocks, and each bank built its own detection and enforcement rather than a shared standard. That leaves room for an agent to slip through a gap between one bank's rules and another's, or for a less-known agent to avoid whatever signature a bank is filtering on. For anyone building or operating an AI agent that touches financial accounts on a user's behalf, the practical read is that a bank's public block announcement does not mean the door is actually shut, and credential-based login to a brokerage or checking account remains a live liability question, not a solved one. Account holders who have granted any agent login access to a bank account should not assume October's announcements removed that exposure.

What to watch next

Watch for Axios or another outlet to name the specific banks, agents, or integration paths behind the continued access, and for any bank to confirm a breach tied to the gap rather than a blocked attempt. Also watch whether Wells Fargo's new Online Access Agreement, which takes a different approach than an outright block and takes effect November 24, becomes a template other banks copy or a cautionary example.

Sources