US banks clamp down on transactional AI tools: Here is what customers need to know

Wall Street is drawing a firm boundary against autonomous technology. Major US financial institutions — including JPMorgan Chase, Bank of America, Morgan Stanley, Citigroup, Wells Fargo, and Charles Schwab — are actively blocking and restricting third-party AI agents from logging into customer accounts and carrying out automated transactions.

The decisive industry crackdown comes as banking security teams detect automated assistants using customer login credentials to access portals, contact customer service call centers, and, in some cases, attempt actions beyond explicit user instructions.

An Axios report said some AI agents attempted to ​take actions that appeared to be independent of user instructions, citing people ​familiar with the matter.

Financial executives warn that allowing software to execute transfers or make purchases introduces unprecedented operational and liability risks.

On October 13, JPMorgan Chase CEO Jamie Dimon publicly addressed the issue, emphasising that allowing unvetted outside tools into internal systems is far too risky for clients. "You've got to make sure customers aren't hurt by it," Dimon noted during a recent earnings call. "You don't want rogue agents, and there could be thousands of them".

Beyond security, banks face a structural threat to customer deposits. Advanced AI agents can automatically optimise customer finances by sweeping idle cash into external accounts offering higher yields. Unrestricted automated fund movements could accelerate deposit flight and destabilize low-cost funding bases across the sector.

To mitigate these exposure points, institutions are overhauling customer agreements and access rules. Wells Fargo and JPMorgan are updating digital service provisions to explicitly clarify liability and restrict unauthorized automated access.

Meanwhile, brokerages such as Charles Schwab are leaning toward restricted, read-only permissions. This middle-ground setup allows third-party tools to review balance and transaction data for planning purposes without holding execution privileges to move money or execute trades.

Industry standards groups and regulators are closely watching the ecosystem to establish formal authentication frameworks tailored for software acting on human behalf. Until secure, industry-wide agent verification protocols exist, major lenders are choosing to shut the door completely on transactional AI access.

What customers need to know 

• Sharing passwords with AI violates bank terms: Handing over primary login credentials to outside AI applications violates standard bank terms of service that mandate direct human authentication. Banks cannot reliably distinguish between a customer-authorized AI and a malicious bot. 
• Transaction execution is being cut off: Third-party AI software will no longer be allowed to execute automated trades, move funds, or make purchases on your behalf through bank portals. 
• Read-only access is becoming the standard: Several brokerages, including Charles Schwab, are limiting third-party access to read-only permissions. You can still use AI tools for balance tracking and financial planning, but not for executing transactions. 
• Terms of service and liability rules are changing: Banks like Wells Fargo and JPMorgan Chase are updating digital service agreements to explicitly outline liability boundaries and strictly prohibit unauthorized automated account access.