RBC and BMO warn against AI agent access to bank accounts

RBC and BMO, two of Canada's largest banks, have warned against AI agents accessing bank accounts, according to reporting by Money.ca. The headline states the position directly: the banks do not want AI agents stepping into account operations.

The caution is not isolated. Once an AI agent can log in on a customer's behalf, check balances or even initiate transfers, the clear boundary of 'the account holder did this' starts to blur. The question banks then have to answer becomes who is responsible for a transaction that no human typed out.

Bank risk frameworks rest on authentication, behavioral patterns and traceability. How credentials are held, how unusual instructions are blocked and who absorbs losses all need settled answers before an agent is let in. The RBC and BMO warning effectively draws a line ahead of those answers.

It also exposes a practical constraint on agentic AI. Model capabilities tend to advance faster than financial institutions can update authorization and audit mechanisms. An agent can click every button in a browser, but a bank may be unwilling to treat those clicks as the account holder's own.

What to watch is whether other banks follow this stance, whether payment and identity systems create dedicated authorization credentials for agents acting on a user's behalf, and whether agent platforms can offer auditable action logs convincing enough to earn bank trust.