Major US financial institutions are tightening restrictions on third-party artificial intelligence (AI) agents attempting to access customer accounts, reflecting growing concerns about security, financial control, and accountability.
The move follows similar restrictions by Amazon on AI-powered shopping agents. Banks including JPMorgan Chase, Citigroup, and Bank of America are limiting unauthorised access by external AI tools, while Charles Schwab has also confirmed that it blocks third-party agents from conducting financial transactions on customers’ behalf.
The developments highlight a growing challenge for the financial industry: how to support AI-powered services without exposing customers or financial institutions to additional risks.
Why Are Banks Restricting AI Agents?
AI agents are designed to perform tasks on behalf of users, potentially including checking account information, moving money, paying bills, and managing financial activities.
However, allowing an external agent to log in using a customer's banking credentials can create security and accountability concerns. If an agent makes an unauthorised transfer, exposes sensitive information, or acts on misleading instructions, determining responsibility can become complicated.
JPMorgan Chase CEO Jamie Dimon has warned about the risks of allowing unverified external agents to access the bank's systems. The bank has said it does not permit third parties to access customer accounts by collecting login credentials and signing in on customers' behalf.
Security, Privacy and Liability at the Centre
1. Protecting Customer Credentials
Giving an AI agent access to banking credentials could expose sensitive financial information. Banks want to prevent external systems from operating inside customer accounts without appropriate safeguards and authorisation.
2. Preventing Unintended Transactions
Autonomous agents can execute tasks quickly, but errors or poorly interpreted instructions could lead to unwanted financial activity. Banks are particularly cautious about allowing third-party systems to initiate transfers or make decisions involving customers' money.
3. Clarifying Responsibility
Another concern is liability. If an AI agent makes a financial mistake, responsibility could be disputed among the customer, the AI provider, and the financial institution.
Charles Schwab CEO Rick Wurster has raised questions about who would be accountable if an external AI agent accessed an account and something went wrong.
Banks Are Not Rejecting AI Entirely
The restrictions do not mean financial institutions are abandoning artificial intelligence. Instead, many are distinguishing between AI tools they can supervise and external agents operating with limited oversight.
JPMorgan, for example, uses AI internally and is developing its own customer-facing applications. Its stated position leaves room for secure data sharing through authorised intermediaries, while restricting unapproved third-party access and account actions.
This distinction is important. Banks can use AI to improve customer service, analyse financial information, and automate internal workflows while maintaining controls over access to sensitive systems.
What This Means for FinTech Companies
The restrictions could influence how fintech platforms, AI developers, and banks design agent-based financial services.
Future systems may need stronger identity verification, limited permissions, transaction limits, explicit customer approvals, and clear audit trails. Secure connections that allow approved data sharing without handing over banking passwords could also become increasingly important.
For AI companies, access to financial accounts may depend less on automating tasks and more on demonstrating that their agents can operate securely, transparently, and within clearly defined permissions.
Conclusion
US banks' restrictions on external AI agents reflect the growing tension between autonomous technology and the need to protect financial accounts. As AI tools can take real-world action, institutions must balance convenience with security, customer control, and accountability.
The next stage of AI adoption in financial services will likely depend on building trusted connections between banks and AI platforms, rather than giving unrestricted access to customer accounts.
