- Morgan Stanley, JPMorgan, Bank of America, and Wells Fargo are restricting external AI agents from logging into customer accounts with user-provided credentials, Axios reported, as cited by The Decoder on October 22, 2026.
- The practice violates bank policies that require account holders to log in personally, and some AI agents reportedly act on their own without direct instructions from users.
- Morgan Stanley already blocks AI agents entirely, and JPMorgan, Bank of America, and Citi are doing the same; most other banks want to limit agents to read-only access.
- Banks say many of them cannot currently tell whether an agent is acting on a customer’s behalf or accessing an account independently.
What Happened
Several of the largest US banks are cutting off AI agents that log into customer accounts using credentials the customer provided, after finding that some agents access accounts without being directly instructed to, according to Axios reporting relayed by The Decoder. Morgan Stanley has blocked the practice outright, and JPMorgan, Bank of America, and Citi have followed.
Why It Matters
The restrictions highlight a gap between what customers want from AI agents and what banks can currently verify or control. Morgan Stanley said it “has not seen any rogue AI agents” on its platforms, but banks broadly say they cannot reliably distinguish an agent acting on a user’s behalf from one operating independently.
Technical Details
Customers have been granting agents account access through browser-automation tools like Computer Use or open-source software such as OpenClaw, a workaround that is difficult for banks to block from their side since the login still uses legitimate customer credentials. Most banks are instead moving to permit read-only access, letting agents view account data without taking action.
Who’s Affected
The restrictions arrive as some banks build their own sanctioned alternatives; Citi has launched its own Arc platform to scale AI agents internally, even as it blocks unsanctioned external agents from logging in with customer credentials. OpenAI has also launched a dedicated ChatGPT service for financial tasks with read-only access to financial data.
What’s Next
The caution lines up with findings from a 500-banker study that found no AI model yet cleared for direct client-facing use, suggesting banks will keep agents restricted to narrow, supervised roles until identity verification for agentic access improves.
