Major U.S. banks are moving to block AI agents from customer accounts after detecting unauthorized third-party agents operating inside those accounts – some of which took actions that appeared to go beyond what users actually asked for. The wave of restrictions, first reported by Axios on October 22 and since corroborated across outlets, marks the firmest line the industry has drawn since autonomous software started showing up at the teller window without an appointment.
In most cases, the banks say, customers handed their own usernames and passwords to an unvetted personal AI agent – behavior that generally violates the bank’s terms of service. Importantly, the agents were not “going rogue” in the breach sense: there are no reports of funds being drained by hacked accounts. The problem is simpler and, in some ways, stranger – people are handing the keys to software nobody fully audited.
Why Banks Are Blocking AI Agents Now
The detection activity has been building for months. In September 2026, Bank of America, Capital One and four other global banks formed a coalition warning that agentic technology – especially AI shopping agents – is moving faster than industry standards and protections can keep up. The logical next step followed quickly: if standards lag, restrict access.
The individual moves have been blunt:
- JPMorgan Chase: CEO Jamie Dimon said on the October 13 earnings call that outside agents will not be allowed inside the bank because “it’s just too risky for clients and for JPMorgan.”
- Morgan Stanley: has begun blocking third-party agents as a precaution, though it said it has not observed rogue agents on its platforms.
- Citi: “We do not permit third-party AI agents or assistants to log into a customer’s account or transact directly within our digital banking platforms.”
- Bank of America, Wells Fargo and Charles Schwab: have all taken steps to limit or block unauthorized third-party agent access, with BofA CEO Brian Moynihan arguing the industry must “put a human being in the loop.”
- PNC: CEO William Demchak called agents transacting inside accounts “incredibly unsafe” and “not going to happen” on the bank’s October 15 earnings call.
Banks Are Writing Incident Plans for the Day After
Detection is only half the playbook. According to Axios, banks are also investing in “day after” scenario planning – the ability to quickly secure or restore customer access if an agent-related incident occurs. Concrete measures include detecting and deleting unauthorized agents inside accounts, updating terms of service, blocking unvetted logins and allowing only read-only access, so agents can view balances without moving money.
Read-only is emerging as the compromise position: most banks have authorized, or are moving toward allowing, agents that can see financial data without taking actions. It is the digital equivalent of letting someone read your menu but not touch the kitchen.
What It Means for the Agentic Finance Push
The restrictions land squarely in the middle of the agentic finance boom. We have already covered why software still cannot hold a bank account of its own and how Singapore put AI agents inside binding bank rules before the U.S. and EU. The banks’ current stance sharpens that picture: innovation is welcome, but only through the front door, with the customer’s explicit consent and a human somewhere in the loop.
One caveat worth stating plainly: Reuters said it could not independently verify the specific incidents described in the Axios report, and Morgan Stanley denied observing rogue agents on its platforms. The broad direction – restrict, monitor, plan – is not in dispute, even if the individual anecdotes remain anonymous-source based.
Sources: Axios (Madison Mills, October 22, 2026), Reuters, USA Today, CBS News, PYMNTS, Bloomberg, and company earnings calls cited above.
