Canada’s Major Banks Restrict Third-Party AI Access to Banking Accounts

Canada’s major banks are drawing a boundary around personal AI agents: outside tools may not directly enter authenticated banking systems or carry out transactions. The shift affects customers who use agents such as Meta Platforms Inc.’s Muse to interact with financial accounts, and reflects concerns about access to sensitive information, actions beyond a customer’s intent and liability when something goes wrong.

Royal Bank of Canada, Canada’s largest lender, told employees Tuesday that it would block third-party personal AI agents from its authenticated digital services. The message followed a security review of the technology and its functions, and was described as consistent with the bank’s approach to responsible AI use.

“After careful consideration and based on what we know today, RBC has made the decision to block third-party personal AI agents from direct access to authenticated RBC digital properties,” said Naim Kazmi, group head of technology and operations, and Erica Nielsen, group head of personal banking, in the memo. They warned that agents accessing banking systems could create unacceptable customer risk by acting beyond what users expect or exposing the bank to liability.

Restrictions extend across major banks

Bank of Montreal, Toronto-Dominion Bank and Bank of Nova Scotia said Wednesday that they were also restricting third-party agents from their digital banking systems. Their statements differ in scope, but point to the same immediate limit: customers should not expect an outside agent to sign in to an account or complete a transaction on their behalf.

BMO spokesperson Jeff Roman said the bank does not allow AI agents to access its digital services or execute transactions, citing customer protection. Toronto-Dominion has decided to block third-party personal agents from directly accessing banking information that requires authentication while it works on safeguards, authorization methods and oversight. Scotiabank said it is limiting third-party agents and will continue watching how concerns develop.

Canadian Imperial Bank of Commerce said it does not support personal AI agents directly opening or accessing accounts. National Bank of Canada said it is assessing both possible customer benefits and risks. Its position leaves open the question of whether it may support some uses later, but it has not announced direct account access for these tools.

RBC spokesperson Jeff Lanthier confirmed the employee memo. He acknowledged that the restriction may disappoint a small share of clients who use these tools, while saying the decision was based on protecting customers and the bank’s systems.

Access and authorization are the sticking points

The distinction in the banks’ statements is between AI tools that can offer information and agents that can act inside an account. Direct access to an authenticated service may let an agent see private financial details or initiate a transaction. If an agent misunderstands an instruction, takes an action the customer did not intend or exposes account information, the customer and bank may face difficult questions about authorization and responsibility.

The statements do not describe the technical controls each bank uses or identify a shared standard for approving agents. Toronto-Dominion’s reference to safeguards, authorization and oversight signals the issues that would need attention before access could change. For now, several banks are choosing to keep outside agents away from account areas that require a customer sign-in.

Several banks said they see potential value in AI while raising concerns about sensitive data. Some have urged customers not to share banking credentials with AI agents. That advice matters because giving an outside tool login details could expose information even where a bank blocks the agent’s direct connection. Customers who have already linked an agent to an account may want to check the bank’s current guidance and review which services or credentials they have shared.

A wider banking concern

Canadian banks are not alone in restricting outside agents. Several U.S. banks have blocked or limited unauthorized tools from accessing customer accounts or moving money. Earlier this month, Charles Schwab Corp. Chief Executive Officer Rick Wurster said the company was actively blocking unauthorized AI agents attempting to move clients’ money.

RBC Chief Executive Officer Dave McKay said in a Bloomberg Television interview last week that the lender was considering blocking Muse from customer information. The bank’s subsequent employee message set out a broader rule covering third-party personal agents, without naming Muse as the only tool affected.

For customers, the practical effect is that an agent may not be able to perform tasks that depend on direct, authenticated access. The banks’ announcements do not say that customers lose access to their accounts through the banks’ own digital services. They do establish that outside tools cannot assume permission to sign in or transact for a customer. Any future change will depend on whether banks can define how an agent is authorized, what it may access and who is accountable for its actions.

For now, customers should use their bank’s own instructions before connecting an AI tool to financial information, and avoid sharing credentials with an agent. The next development to watch is whether banks introduce approved ways for agents to assist customers while keeping account access and transaction authority under clear control.