TD temporarily restricts external AI agents from client accounts

TD, a bank, restricts personal AI agents from other companies. They cannot get direct access to banking information where a client must sign in. TD tells that the measure is temporary while it studies the risks. It gave no date for the end of the measure. Vlad Shpilsky, a TD group head, tells that privacy, security and client control come first.

How it works

An AI agent is software that acts for a person, for example to pay a bill. TD restricts such agents where they must go through the sign in check. Links that use an API are not affected. An API is a set of rules that lets programs share data. These links let a client share account data without the TD username or password.

The numbers

TD gave no numbers. It gave no date, no count of blocked agents and no test results. TD tells that it uses agentic AI in the bank with strong governance and controls. Agentic AI is AI that acts on its own.

What is not known

TD does not tell how it finds an agent, or when the restriction ends. It tells that it will study how agents and industry standards change. It tells clients to keep their credentials private and to know the tasks an agent can do.

Insight: Banks close the front door and keep the API door open

Citigroup tells that it does not let third party agents use accounts. TD restricts agents too, but keeps API links open. Economist Torsten Slok tells that agents can move cash from accounts with 0.1% interest to accounts with 3.3% to 5%. Perhaps this is a toll gate: the bank, not the agent, picks who gets in and sets the terms. Agent builders must plan for that. Agent builders must plan for that.