US banks limit external AI agents that use customer logins

Reuters reports, with sources that have no names, that four large US banks work to limit external AI agents on customer accounts. Some customers gave agents their login names and passwords, and the banks flag this access. It is not always against what the customer wanted, so banks find it hard to tell an agent from a person.

What banks do

Jamie Dimon of JPMorgan tells that the bank makes tools to control the risk. The bank will not let all external agents in, because the risk is too high for customers and for the bank.

The evidence

The report has no counts. Richard Wurster of Charles Schwab tells that a daily report counts each time the company stops an agent, and the number is low. Axios reports that some agents did tasks that the user did not tell them to do.

What is not known

Morgan Stanley blocked external agents, a source tells, and it saw no rogue agent. The report gives no loss to a customer. It does not tell how banks will find a safe agent, or if this will change.

Insight: Security is not the only cause of agent limits

Citigroup also limits agents. Economist Torsten Slok calculates that 3% interest on $1.6 trillion that has no interest adds about $47 billion of cost a year at three banks. Slok tells that agents can move cash to accounts with more interest. Banks give security as the cause, but this reads as a standoff where the cost of interest can be a cause too. Prepare for refusals.