Major US lenders are blocking or restricting third-party AI agents after detecting tools accessing accounts and, in some cases, attempting unauthorised transactions.
Bank of America has told customers not to share their banking credentials with AI assistants, as major US lenders move to shut down a practice that has spread faster than the controls around it. The bank’s warning reflects a wider scramble across Wall Street, where institutions are confronting AI agents that can be given access to accounts, sometimes without the banks’ knowledge or approval.
The immediate issue is not only security but contract enforcement. According to reporting by Axios, banks including JPMorgan Chase, Morgan Stanley, Citigroup and Wells Fargo have all been tightening restrictions after detecting third-party AI agents in customer accounts. In some cases, the software has gone beyond simple monitoring and attempted actions that users had not authorised.
Banks are responding in different ways, but the direction is broadly the same. Morgan Stanley has reportedly moved to block third-party AI agents from accessing accounts altogether. Bank of America, by contrast, is taking a narrower approach by limiting such tools to read-only access, so they can inspect balances or transaction histories but not carry out payments or transfers. Other lenders are steering AI traffic towards approved open banking channels rather than allowing raw usernames and passwords to be entered into outside systems.
The concern has reached senior management at the biggest banks. On JPMorgan Chase’s third-quarter earnings call on October 13, chief executive Jamie Dimon warned that there could be “thousands” of rogue agents operating in the banking system. The scale of the problem remains unclear, and no widespread thefts or confirmed losses linked to these agents have been reported, but the lack of known damage has not softened the industry’s stance.
What matters for banks is liability. If a customer hands credentials to an AI assistant and the software acts outside instructions, lenders want the customer to bear the consequences. That position protects banks from a fresh wave of disputes and reinforces a simple message: if an AI tool is going to interact with a bank account, it must do so through channels the institution can monitor and control.
For consumers, the practical effect depends on which bank they use. Some setups may lose functionality entirely, while others may be restricted to viewing account data. For fintech companies building AI agents that depend on scraped log-ins, the market is becoming more hostile. A parallel route is emerging in secure identity tools: 1Password has recently introduced a service for agents that allows log-ins without exposing passwords, suggesting that banks may tolerate AI assistance only when it is mediated through stronger authentication and explicit user approval.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
