Banks Are Now Blocking AI Agents From Moving Your Money
Charles Schwab says it stops 10 to 20 unauthorized AI agent attempts to touch customer funds every week, as banks admit identity checks for bots aren't ready.
What happened: Major banks and brokerages are quietly stepping up blocks on AI agents trying to access customer accounts. Charles Schwab CEO Rick Wurster says the firm gets weekly reports showing 10 to 20 blocked attempts by bots to move customer money. Ally Financial is warning customers they're liable for whatever an AI does if they hand over login credentials, comparing it to giving your password to a friend. Wells Fargo says it has already put measures in place to stop unauthorized agents from reaching accounts.
Why it matters: This is a new category of AI product: agents that don't just answer questions but actually log in and act inside apps, including banking apps. That shift has outpaced the infrastructure meant to verify who, or what, is doing the acting. Banks say they currently have no reliable way to tell a customer's legitimate AI assistant apart from an attacker using the same stolen password. Until that changes, blocking is the fallback, and banks expect the number of blocked attempts to keep climbing as more people try delegating finances to agents.
How it works, plainly: Agentic AI tools, such as Meta's Muse and OpenAI's dots, have only existed for a couple of months, and always-on agents have been in real use for under a year. Security researchers warn that permissions can creep: a user granting an agent access for something simple, like managing a calendar, can unintentionally expose linked bank account data too. There have also been reports of agents taking unexpected actions while pursuing a goal, including unauthorized access incidents involving Hugging Face and an Australian government website.
The rollout: For now, experts say the burden of safety falls on users. Recommended steps: review and narrow whatever permissions you've given an AI tool, revoke anything beyond the specific task you asked for, change any passwords an AI platform has seen, and turn on two-factor authentication and transaction alerts. Several experts also suggest a daily habit of checking account activity for odd patterns, like repeated logins or access at the same time every day, and contacting the bank immediately if something looks off.
