The largest bank in Canada just put a "Do Not Enter" sign on Meta's AI agent, and every enterprise CISO is watching.

The Summary

The Signal

RBC isn't worried about a chatbot. They're worried about an agent that can move, act, and potentially obligate them without human oversight. McKay framed it as a liability question: "You have to protect that consumer and then we're liable for that as an organization."

That word, liable, is doing heavy lifting. When an AI agent accesses customer financial data, who owns the outcome? If Muse pulls transaction history to help a user dispute a charge, and gets it wrong, is that Meta's problem or RBC's? If it recommends a financial product based on incomplete data, who pays when it goes sideways?

"You have to protect that consumer and then we're liable for that as an organization."

Banks live and die by regulatory compliance. Every customer interaction is logged, audited, defensible. AI agents break that model. They operate in natural language, make probabilistic decisions, and leave messy audit trails. For a bank like RBC, which manages $1.7 trillion in assets across 17 million clients, letting a third-party agent roam free inside customer accounts isn't innovation. It's exposure.

Meta's Muse has become the breakout consumer AI agent of 2026, the first to hit mainstream adoption outside of narrow productivity use cases. It books travel, manages subscriptions, answers emails, and increasingly, handles financial tasks. That popularity is exactly what makes it dangerous for enterprises.

Here's what RBC is really saying:

  • We don't control Muse's training data or decision-making logic
  • We can't audit what it does with customer information after it leaves our systems
  • If something goes wrong, regulators will hold us responsible, not Meta

This isn't about trust in Meta's security. It's about control. Banks spent decades building walled gardens around customer data. Now agents want the keys to every door, and they're not asking permission, they're asking forgiveness.

The Implication

If RBC actually blocks Muse, expect other banks to follow. Not because they're anti-AI, but because the liability math doesn't work yet. The agent economy can't scale in regulated industries without clear answers on accountability, auditability, and who pays when the agent screws up.

Watch for two things: banks building their own approved agent ecosystems with strict API governance, and pressure on Meta to offer enterprise-grade agent contracts with actual liability coverage. The Wild West phase of AI agents just hit its first border checkpoint.

Sources

Bloomberg Tech