RBC CEO says bank is considering blocking Meta’s AI agent Muse from accessing customer data

Royal Bank of Canada is considering restricting access by Meta Platform Inc.’s AI agent Muse to the bank’s customer information, according to comments made by Dave McKay, the bank’s president and CEO, in an interview posted on Yahoo Finance on Oct. 20, 2026.

McKay said the bank needs to protect customer data, framing the decision as a question of controls and safeguards rather than a judgment on AI in general. The comments indicate RBC is actively evaluating whether an external or third-party AI tool could be granted permissions that would allow it to view or use customer information.

The interview described Muse as an “artificial intelligence agent” associated with Meta. In the context of a regulated financial institution, the key issue is not whether the technology can produce answers, but what data it can access and how that data is governed inside the bank’s systems.

While the comments point to a potential block or restriction, the interview did not provide details on the mechanism, such as whether RBC would prevent Muse from connecting to internal data sources, limit the scope of what prompts can include, or apply a new approval process for specific AI use cases. It also did not specify whether the evaluation is limited to one department, one project, or a broader enterprise policy.

The episode reflects a broader enterprise pattern as companies experiment with AI assistants and agents. As organizations move from chat-based tools to systems that can perform tasks, the permission model becomes central: which datasets are accessible, what logging and auditing exist, and how permissions are revoked when risks are identified.

For banks, those questions are amplified by regulatory expectations around privacy, data handling, and operational resilience. Even when a vendor promises strong security measures, institutions often decide whether any external system should be allowed to view customer information directly, especially for tools described as capable of taking actions rather than just generating text.

A related complication is that “AI agents” are often implemented through combinations of software components, integrations, and cloud services, and those implementations can vary from one customer or partner deployment to another. Without further disclosure from RBC or Meta, it is unclear what, if any, current integration Muse has with RBC systems, and what data categories were at issue in the CEO’s remarks.

Going forward, investors and industry observers will likely watch whether RBC clarifies the scope of its evaluation and whether it adopts a formal AI governance policy covering third-party agents. They will also watch how Meta responds to public scrutiny around data access and controls, especially as more large organizations push for tighter constraints on what AI tools can see.