Why Major Banks Are Restricting AI Agent Access and Building Safer Alternatives

Personal AI assistants are becoming more capable, creating new opportunities and risks for the banking industry. Tools from major technology companies can now compare options, make recommendations and potentially handle financial tasks on behalf of consumers.

That development is prompting large banks to reconsider how outside AI agents interact with customer accounts. During recent third-quarter earnings discussions, analysts questioned bank executives about two major issues: whether AI agents could encourage customers to move deposits in search of higher yields, and whether autonomous software should be allowed to execute banking transactions.

Bank leaders generally downplayed the immediate threat to deposits. However, several institutions made it clear that they are blocking or limiting access from external AI agents until stronger security, authentication and liability standards are in place.

JPMorgan Warns Customers About External AI Agents

JPMorgan Chase provided one of the clearest statements on the issue. Chairman and Chief Executive Officer Jamie Dimon warned customers about sharing bank credentials with outside AI tools.

According to Dimon, customers could face the consequences if an external agent misuses their information or makes an unauthorized decision. He said JPMorgan is developing its own tools but is not prepared to allow unapproved agents into the bank’s systems because of the potential risks to customers and the institution.

The message from JPMorgan reflects a wider concern across the financial sector: an AI agent may act on a customer’s behalf, but it may not always understand the customer’s intent, financial obligations or risk tolerance.

Could AI Agents Trigger Deposit Switching?

Earlier this year, banking analysts warned that a simple instruction such as “optimize my idle cash” could force banks to compete more aggressively for deposits. An AI assistant could theoretically examine a customer’s accounts, compare interest rates and transfer funds to a higher-yielding provider.

For now, bank executives say that scenario is not creating a significant near-term threat.

Bank of America Chief Executive Officer Brian Moynihan said the usefulness of automated deposit optimization depends on the type of account involved. An AI tool may identify a higher interest rate, but it may not know when a customer needs money for an upcoming payment or when moving funds could create an overdraft or insufficient-funds issue.

Customers with savings or investment balances are already seeking better returns through existing bank services, Moynihan noted. Transaction accounts, however, serve purposes beyond earning interest.

Wells Fargo Chief Executive Officer Charles Scharf also questioned whether AI agents would quickly lead to widespread movement of deposits. Consumers can already compare rates and transfer money online, he said, but many people retain funds with a bank for convenience, trust, payment services and broader financial relationships.

Scharf acknowledged that AI agents will become part of how consumers interact with financial providers. However, he said the expectation that they will soon cause mass transfers to the highest-rate accounts is exaggerated.

U.S. Bancorp Chief Executive Officer Gunjan Kedia said the bank has not seen third-party AI tools attempting to move customer cash. The process remains complicated, and the institution does not expect automated cash management by outside agents to have a major short-term impact.

Kedia added that U.S. Bancorp already offers digital tools that help customers compare deposit pricing and automatically manage balances when appropriate.

Banks Are Developing Their Own Cash Management Tools

Several financial institutions are creating internal services that provide some of the benefits associated with AI-driven deposit optimization while keeping activity inside the bank’s controlled environment.

• Citibank introduced Citi Sky for wealth management customers. The service helps clients evaluate options for funds released when a certificate of deposit matures.
• Chase has been testing Smart Cash, a service designed to move money among eligible accounts within the bank to help customers improve returns.

These services allow banks to respond to customer demand for automated financial management without giving unverified outside agents unrestricted access to account systems.

Could Agentic AI Accelerate a Bank Run?

Analysts are also examining whether autonomous AI could increase liquidity risk. If an agent monitored news, market conditions or social media and then moved funds for thousands of customers at the same time, deposit outflows could potentially occur faster than in the past.

During a JPMorgan earnings discussion, an analyst asked whether banks and regulators should change their approach to liquidity stress testing because of agent technology.

JPMorgan Chief Financial Officer Jeremy Barnum expressed skepticism that AI alone would create a new type of bank run. He argued that the collapse of Silicon Valley Bank was driven primarily by a concentrated and undiversified depositor base responding to a shared concern.

Barnum said similar speed could have occurred with older communication technologies. Still, he acknowledged that the growing use of bots and autonomous systems could create new risks that regulators and banks need to prepare for.

Dimon also said financial institutions and regulators should consider the possibility that automated agents could make mistakes or cause damage at scale. His broader message was that banks must maintain strong liquidity, risk controls and customer confidence.

Why Banks Are Blocking Unapproved AI Access

PNC Chairman and Chief Executive Officer William Demchak described autonomous agents conducting banking transactions as unsafe under current conditions.

Demchak said any future system would require clear liability rules, secure connections, strong authentication and human oversight. He also rejected the idea that an outside program should be able to transact on a customer’s behalf without the customer knowing what is happening.

JPMorgan has similarly raised concerns about large numbers of unauthorized or poorly controlled agents operating inside financial systems. Banks are particularly wary of screen scraping, a practice in which a third-party service uses customer credentials to access account information.

Executives say screen scraping is difficult to secure because customers may not know exactly what an automated tool is doing, what information it is collecting or how it is storing login details.

What Banks Want From AI Agent Partnerships

Although banks are restricting uncontrolled access, they are not rejecting agentic AI altogether. Instead, many institutions are working toward secure partnerships with technology companies, data aggregators and AI developers.

Bank of America is exploring ways to work with AI providers while keeping customers in control of their financial decisions. The bank’s goal is to support automated assistance without allowing an agent to move money or change account settings without proper authorization.

JPMorgan executives have also said they are developing a safer framework for AI-enabled activity. The institution wants to prevent credential sharing while creating approved methods for agents to interact with its systems.

American Express is taking a similar approach. Chief Executive Officer Stephen Squeri said the company is working with AI companies, including OpenAI, to integrate its card services into emerging AI experiences.

The company’s strategy includes helping customers receive relevant recommendations, complete secure payments, access card benefits and obtain support after a transaction. Squeri emphasized that financial companies must adopt AI in a way that reflects how customers want to use the technology.

The Future of AI in Banking

The banking industry is moving toward a controlled model of agentic AI. Banks want to offer automated recommendations, cash management and customer support, but they do not want unverified software accessing accounts through shared credentials or uncontrolled connections.

The biggest challenges will involve authentication, data privacy, fraud prevention, liability and human oversight. Regulators may also need to establish rules governing responsibility when an AI agent makes an error or causes a financial loss.

For consumers, the safest approach is to avoid sharing bank passwords with unverified AI tools. Until secure industry standards become widely available, customers should carefully review permissions, monitor account activity and confirm any automated transfers or financial decisions.

AI agents are likely to become an important part of digital banking. However, major banks are signaling that this transition will happen through approved systems, secure partnerships and customer-controlled authorization rather than unrestricted access.

Source: TheFinancialBrand.com