Bank of Montreal’s CEO Darryl White has voiced concerns regarding the reliability of artificial intelligence agents when it comes to handling personal banking information.
During his address at the Toronto Global Forum on Wednesday, White highlighted his apprehension about the rapid advancement of AI outpacing regulations and protective measures. His comments followed those made by RBC’s CEO on the previous day regarding the possibility of restricting Meta’s widely-used Muse AI agent from accessing customer data at his institution.
White stated, “For 25 years, we have become accustomed to an e-commerce system.” He noted that this familiarity has led some people to be overly willing to “enter all my personal information anywhere, into any agent.”
He urged, “Do not disclose your banking username and password. Avoid giving out your credit card, social insurance number, or PIN. These details are not safeguarded.”
“I believe there is still much work to be done in educating the public,” White added.
In recent years, Canada’s major banks have poured billions into artificial intelligence. In August, White proudly mentioned that BMO’s insurance sector is utilizing AI to make underwriting decisions in as little as 10 seconds, a significant reduction from the former industry benchmark of 28 days.
With the consent of clients, AI tools like Meta’s Muse and OpenAI’s ChatGPT are capable of linking to bank accounts and credit cards to track balances and transaction histories, while also providing advice on spending patterns. Typically, making purchases requires human verification.
On Tuesday, RBC’s CEO Dave McKay indicated that his bank is contemplating the option of preventing Muse from accessing client data.
“It’s definitely something I’m considering,” he remarked during a television interview with Bloomberg News.
He added, “We won’t rush into a decision, but it’s crucial to balance consumer needs with the privacy and security of the system.”
Recently, the heads of JPMorgan Chase & Co. and Charles Schwab announced their intentions to restrict or limit external AI agents from utilizing their clients’ data.

