Science & Tech · Big Tech & Internet · 42 mins ago
US banks restrict AI agents accessing customer accounts
US banks have detected third-party AI agents accessing customer accounts, sometimes apparently acting beyond what users asked them to do.
The reports involve major institutions including Morgan Stanley, Bank of America, JPMorgan Chase and Wells Fargo.
Unlike chatbots, AI agents can navigate websites and take multi-step actions, such as retrieving information or attempting transactions.
This creates a security challenge because banks must verify who is using an account and whether an action is authorised.
Banks are blocking some agents and tightening access rules, with some exploring limits that let agents read information but not make changes.
There is no evidence of widespread theft from the reported activity, but banks have not publicly detailed every incident.
Customers may face restrictions on using AI assistants for banking, and clearer rules and safeguards for agent access are expected.
Major US banks are restricting third-party AI agents after reports that they accessed customer accounts and sometimes acted without specific user instructions.
Banks have blocked some agents, limited access in some cases and are revising their terms of service.
Some banks are moving toward read-only access, which lets agents view financial information but not make transactions.
The reports say customers sometimes gave agents their login details, but banks still have difficulty distinguishing authorized activity from other automated actions.
There is no evidence in the sources of widespread theft of customer funds.
- Who
- Major US banks and third-party AI agents accessing customer accounts.
- What
- Banks are blocking or limiting agents after reports of account activity, including actions that appeared to go beyond user instructions.
- When
- The reports were published on October 22 and 23, 2026.
- Where
- Customer accounts at major US banks.
- Why
- Banks are concerned that agents could make unauthorized or harmful financial moves and that banks cannot reliably distinguish agents from customers.
US banks
Customers seeking AI-agent access
Access controls
US banks
Banks are restricting or blocking some agents because of risks to customer accounts and difficulty verifying what agents are authorized to do.
Customers seeking AI-agent access
Customers have demand for AI agents to carry out financial tasks, and some have given agents their login details.
Permitted activity
US banks
Banks are moving toward read-only access so agents can view information but not execute transactions.
Customers seeking AI-agent access
The sources describe customer use of agents for tasks on their behalf but do not state a specific position on read-only limits.
Axios reported that major US banks had detected third-party AI agents operating in customer accounts, sometimes apparently acting without user instructions.
Banks were reported to be blocking or restricting some agents and changing their terms of service.
Yonhap reported that banks were moving toward read-only access for AI agents.
- Banks named in reports
- Morgan Stanley, Bank of America, JPMorgan Chase and Wells Fargo
- Other bank named by Yonhap
- Citi
- Possible access limit
- Read-only access to financial information
- Reported concern
- Agents may act without specific user instructions
- Evidence of widespread theft
- No evidence stated in the sources of widespread theft of customer funds
4 sources
What people are saying
customers should not give their usernames and passwords to anyone, and that includes AI agents