US Banks Blocking AI Agents in Accounts

Major U.S. banks including Morgan Stanley, JPMorgan, Bank of America, Wells Fargo and Citi are actively blocking external AI agents that log in with customer credentials, arguing such access violates rules that require account holders to sign in personally and creates unacceptable security risk. Banks say some AI agents act autonomously without direct user instruction, and many institutions cannot reliably distinguish agent-driven activity from human actions. Morgan Stanley has instituted a full block on AI agents and notes it has not seen "rogue" agents on its platforms, while other banks are moving toward similar restrictions. The prevailing stance among banks is to permit only read-only access to financial data, not transaction capability, because customers want AI assistance but the risks of unauthorized transfers or actions are too high.

Agents commonly gain access by operating through a customer’s browser using tools like Computer Use or open-source software such as OpenClaw, which makes provider-side blocking difficult to enforce. As a counterpoint, OpenAI has launched a U.S. ChatGPT service tailored for financial tasks that provides read-only connectivity to accounts, reflecting industry momentum toward limited, controlled integrations rather than full agent-driven account control. Banks are tightening controls and exploring safer, constrained ways to let customers use AI without exposing accounts to autonomous agent behavior.