The password is the problem

Read RBC's notice closely and the objection is not to advice, summaries or spending analysis. It is to the login. RBC's stated concern is how third-party personal AI agents access authenticated accounts. Its notice warns that such tools may require access to customers' online accounts and may store their passwords. That creates security and authorization questions when software acts on a customer's behalf. RBC has chosen to restrict direct agent access rather than rely on those arrangements.

RBC's notice also points customers to its Digital Banking Security Guarantee, which describes circumstances in which reimbursement is excluded, including password sharing and transactions the customer authorizes or contributes to. The notice does not determine how those provisions would apply to every agent-related loss or dispute. The bank's separate policy decision is to block direct access by third-party personal AI agents.

A decision with the edges left blank

What RBC has not published matters as much as what it has. There is no list of which tools count as a "third-party personal AI agent," no per-product ruling a customer can check, and no date when existing connections stop working. The notice does not define the category through a published product list or technical criteria. Customers using third-party personal AI agents therefore cannot determine from the notice alone whether a particular product is covered or when an existing connection will stop working.

According to The Globe and Mail's reporting, OSFI Superintendent Peter Routledge said banks could make their own decisions about allowing agent access, provided they acted responsibly. That leaves room for different bank policies under the current approach described in the report; it does not mean agent access falls outside regulatory oversight or that future guidance is ruled out.

The pattern behind one bank

RBC is early in Canada and late in a wider sequence. The clearest financial-sector comparison in the reporting is Charles Schwab, whose chief executive said the firm was proactively blocking unauthorized agents attempting to move client funds. RBC and Schwab illustrate a narrower pattern: financial institutions are restricting some forms of unauthorized agent access where account control and transaction authorization are central concerns. That does not establish a common policy across banks or other industries.

That leaves agent makers with a narrow path. Stored-password access is the thing being closed off. RBC directs customers to its official website and mobile app, its own digital tools and customer service. The notice does not describe an approved direct-access route for third-party personal AI agents.

Sources

  1. [1] RBC, "Understanding third-party AI agents and your RBC banking," October 27, 2026 (public notice; page metadata published October 28, 2026 UTC; the blocking decision, the password and systemic-risk reasoning, the stop-working warning, the continuing routes, and the Security Guarantee terms are all from this page, read in full this run)Read source
  2. [2] Bloomberg Law, "RBC to Block AI Agents, Calling Out 'Unacceptable' Client Risk," October 27, 2026 (staff memo seen by Bloomberg; Kazmi and Nielsen as senders; full security review)Read source
  3. [3] Briefs Finance, "Royal Bank of Canada blocks third-party AI agents from its systems," October 27, 2026 (syndicated summary of Bloomberg and Globe reporting: Lanthier confirmation and "small portion of our clients" quote, Schwab's Wurster comparison, McKay's Bloomberg Television comments the previous week)Read source
  4. [4] Madrobot, "RBC blocks AI agents like Meta's Muse from online banking," October 28, 2026 (synthesis of Globe and Bloomberg reporting: first-Canadian-bank characterisation attributed to the Globe, OSFI's Routledge position, the absence of a named-product list)Read source