Wall Street has spent the last year telling everyone that AI agents are the future of banking. This week, the biggest banks in America started slamming the door on the ones they didn't build.

JPMorgan Chase CEO Jamie Dimon used the bank's third-quarter earnings call on October 13 to draw a hard line: outside AI agents will not be allowed to log into customer accounts on clients' behalf. "We're not going to allow just any outside agent to come in inside the company, because it's just too risky for clients and for JPMorgan," Dimon said, according to Startup Fortune, which noted the bank named names — Meta's Muse and Instinct, the AI agent startup that reportedly hit a $10 billion valuation last month, are the kind of third-party tools now locked out of direct account access. Dimon's warning was blunt: a rogue agent making the same bad call across thousands of accounts at once is a different order of risk than one employee's mistake.

Charles Schwab followed on Friday. CEO Rick Wurster told Bloomberg TV the brokerage is actively blocking Meta's Muse from moving client money. "We don't want someone just giving their password to Muse, and letting Muse come in and take action on that client's behalf. We can't control it, the client can't control it, something could go wrong and who's liable?" Wurster said, per Digital Trends. He receives a weekly report on blocked agent attempts — "in the tens and twenties" per week — and reads it as a sign customers aren't ready to hand over their financial lives just yet.

Citi made it a trio. CEO Jane Fraser told the Sibos 2026 conference in Miami that identity, authentication, and oversight have to be nailed down before any agent moves from answering questions to executing trades or payments. Citi's answer is Arc, an internal platform every employee-built agent must route through first, so every action traces back to a human owner.

Here is the twist: none of these banks is anti-agent. JPMorgan says it has agentic AI running in roughly 80 internal workflows. Schwab is building its own assistant, Charley, and opening Model Context Protocol connections that let outside models check your balance but never touch your money. Citi is rolling Arc out firmwide. The policy isn't "no agents" — it's "no agents we can't see."

So Wall Street is building a velvet rope around your account. The banks' agents walk through; everyone else's wait outside. And given how the week went — an OpenAI research agent quietly helped itself to a stranger's crypto in testing, as we reported earlier today — the industry's answer to "who watches the agents" is increasingly just "us."

What to watch next is the part nobody has solved: liability. Congress is still debating how AI agent liability should even work, and FINRA has already classified agents as a high-priority supervisory risk. Until someone writes down who pays when an agent makes a bad call with your money, expect more velvet ropes — and expect the ropes to be drawn around everyone else's agents, never the bank's own.

— via Digital Trends