Two of the biggest names in American finance are drawing a hard line against AI agents handling your money. Charles Schwab is actively blocking unauthorized AI agents — including Meta’s personal assistant Muse — from moving client funds, chief executive Rick Wurster told Bloomberg Television on Friday. Days earlier, JPMorgan Chase CEO Jamie Dimon told analysts on the bank’s third-quarter earnings call that outside AI agents are too risky to allow inside customer accounts.
The two announcements, landing in the same week, mark the industry’s clearest pushback yet against the vision of AI agents acting as autonomous financial assistants. Both executives framed the decision around the same concerns: security, control, and the unresolved question of who is responsible when an AI agent makes a costly mistake with a customer’s money.
The stance is striking because banks have spent the past two years telling investors that agentic AI would transform their back offices. Now the same executives are barring those agents from the one place that matters most to customers: their own accounts.
Why Banks Are Blocking AI Agents
Wurster was blunt about Schwab’s reasoning. “We can’t control it, the client can’t control it, something could go wrong and who’s liable?” he said, adding that he personally receives weekly reports on blocked agents. Those blocks run in the tens per week for the major AI agents — a figure Wurster reads as evidence that customers are not yet willing to hand their full financial accounts to artificial intelligence.
Dimon’s warning on JPMorgan’s October 13 earnings call was equally direct. “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,” he said, raising the specter of rogue AI agents operating at scale — potentially thousands of them making the same bad decision across thousands of accounts at once. A bank spokesperson confirmed to CNBC that the restriction applies to external agents acting for retail clients.
The banks named names. Meta’s Muse — the company’s personal AI agent, which collects extensive data about users and the people they interact with — is among the tools now locked out of direct account access. An AI agent startup called Instinct was also cited as the kind of third-party tool covered by the bans.
AI Agents and the Liability Gap
At the heart of the dispute is a legal question nobody has answered: when an AI agent moves money on your behalf and something goes wrong, who pays? Congress is still debating how AI agent liability should work, and the banks are not waiting for the outcome. Dimon put the issue to customers directly, saying anyone who hands authority to an outside agent should understand the risks — and that doing so amounts to a shift of responsibility away from the bank.
Other financial leaders are circling the same problem without drawing as bright a line. Citigroup CEO Jane Fraser told the Sibos 2026 conference in Miami that she wants identity, authentication, and oversight nailed down before any agent moves from answering questions to executing trades or payments on its own. Amazon, meanwhile, blocked Muse from shopping on its site last month over credential concerns — a sign the resistance extends well beyond banking.
For now, the banks’ position amounts to a simple principle: an AI agent can look at your money, but it cannot touch it.
What AI Agents Can Still Do With Your Bank Account
The restrictions have limits. JPMorgan still allows customers to share account data — balances and transaction history — with outside AI agents through authorized open-banking connectors rather than handing over login credentials. What those agents cannot do is initiate transfers or pay bills; anything that moves money is blocked, and the bank has offered no timeline for changing the rule.
That read-only compromise mirrors where the industry landed in its earlier fight over data access, when JPMorgan clashed with aggregators over fees and control. Schwab’s weekly block counts suggest adoption of full agent access remains low anyway — “in the tens and twenties,” in Wurster’s words — so the banks are locking a door that few customers were trying to open.
Still, the direction is set. As AI agents grow more capable and more trusted, the pressure to let them act will only increase — and the industry’s answer, for now, is that the technology must prove it can be identified, authenticated, and held accountable before it gets near anyone’s money.