RBC, JPMorgan, Citi and UBS Move to Block Third-Party AI Agents from Customer Accounts
Major banks in Canada, the US and Switzerland are restricting third-party AI agents such as Meta's Muse from accessing customer accounts, citing fraud risk and unclear liability.
A wave of major banks is blocking or restricting third-party AI agents from accessing customer accounts. In Canada, RBC, TD, BMO and CIBC have moved to block them, with Scotiabank imposing limits; RBC issued a customer memo on October 28. In the United States, JPMorgan, Citi, Bank of America, Wells Fargo and Morgan Stanley have taken similar positions, while UBS confirmed it does not permit external AI models in client transactions. The catalyst is agentic tools such as Meta's Muse, which can act on accounts on a user's behalf.
Banks cite the risk of agents acting beyond their instructions, fraud exposure, and unclear liability when something goes wrong. Canada's regulator OSFI has left policy to individual banks while tightening its AI guidance. No bank has reported an actual incident; Morgan Stanley said it has not yet seen rogue agents in the wild.
The decisions effectively start defining what a "trusted" bank-approved AI agent looks like: explicit consent, audit trails, guardrails and verified identity. Banks that block outside agents still need internal automation, which is accelerating demand for enterprise AI agents deployed inside the bank's own environment.