Royal Bank of Canada’s head of AI says the lender is concerned about AI agents accessing customer accounts and moving money on their behalf, adding that the firm will not let the likes of Meta’s Muse access its systems without sufficient safeguards in place. 

In an interview with The Logic, Bruce Ross said RBC also needed to ensure it could prevent unauthorized activity by AI agents and establish who bears responsibility when something goes wrong. Earlier this week, RBC CEO Dave McKay said the bank is considering blocking Muse from accessing customer information, though it is yet to make a final decision. RBC spokesperson Briana D’Archi declined to answer questions about what a potential ban would look like or what measures the bank could use to prevent agents from accessing customer accounts.

JPMorgan Chase CEO Jamie Dimon raised similar concerns during the lender’s third quarter earnings call last week, warning customers about the risks of granting AI agents access to their accounts. The U.S. bank is also developing protections against rogue AI agents.

Concerns over AI agents are also attracting the attention of international bodies. The Financial Stability Board, with support from Canada’s banking regulator, the Office of the Superintendent of Financial Institutions, has urged financial institutions to strengthen oversight of AI systems, including verifying AI agents’ identities and restricting their access to sensitive information.

The challenge is particularly complicated in Canada, which remains a global outlier among similar economies when it comes to open banking. Unlike in the U.S., the U.K. and the EU, where banks are broadly mandated to let consumers securely share financial data with approved third parties, Canada still lacks comprehensive open banking rules. Even proposed regulations focus only on data access, rather than the authorization of payments.

With open banking rules in Canada still years behind those in other countries, many fintechs rely on screen scraping, a practice in which customers provide their banking credentials to third-party services that log into their accounts and extract information from them. The practice violates the terms of service of some of the big Canadian banks, but lenders have generally been reluctant to block it outright because many customers rely on financial apps that use it.

RBC’s caution regarding AI agents developed by other firms contrasts with its plans to develop and deploy artificial intelligence internally. The bank is marking the 10th anniversary of Borealis, its in-house AI research institute, and is expanding its use of AI across credit decisions, wealth management and investment banking.

Founded in September 2016, Borealis began as a small research operation at Toronto’s MaRS Discovery District, where a small group of researchers worked to advance machine-learning technology and explore its applications in banking. 

Over the past decade, RBC has invested billions of dollars in data and AI initiatives, Ross said. Those investments have helped the institute develop proprietary AI models, including ATOM, a machine-learning foundation model that enables the bank to personalize offers to customers, better understand credit needs and inform lending decisions. Ross said ATOM drives much of the economic value RBC currently generates from AI. “We didn’t just drop this Ferrari on the table in front of the businesses and functional leaders. We built it with them, and so now it’s there to drive,” he said.

Despite RBC’s advances in AI, the bank has no immediate plans to sell its proprietary AI technologies to other financial institutions or businesses, Ross said. While the bank has considered commercializing its technology to generate additional revenue, it sees greater value in using it to improve operations and expand customer services, he added. 

RBC’s AI investments have helped it hold third place in the Evident AI bank ranking for three consecutive years, behind only JPMorgan Chase and Capital One. Daniel Shackleford Capel, Evident’s head of banking, credits Borealis as a key reason behind the bank’s success.

Ross has his sights set on the top spot, saying that a number of other banks are playing catch-up when it comes to AI.

RBC said at its March 2025 investor day that it expects AI to generate between $700 million and $1 billion in enterprise value by fiscal 2027. Ross said the bank remains confident it will meet that target, although he declined to quantify how much of it has already been achieved.

Ross said Borealis would continue to focus on using AI to protect RBC’s customers and systems, while also exploring how advances in the technology, and in fields such as quantum computing, could benefit the company and its customers in the coming years. He cautioned that the speed of technological development makes it increasingly difficult to predict the future. “I think if you’re predicting three years out, you’re doing really well,” he said.