Major US banks are tightening restrictions on third-party artificial intelligence agents that access customer accounts using login credentials supplied by users. The restrictions highlight growing concerns over security, transaction authority, and accountability as AI agents become increasingly capable of carrying out financial tasks.
The restrictions also expose a gap in US open banking regulation. A federal rule designed to replace password-based account access with secure data-sharing interfaces remains stalled in court, leaving banks and technology companies without a consistent framework for managing third-party AI access.
Banks Raise Concerns Over AI Agent Activity
Reports from Axios and Reuters indicate that major US financial institutions have been detecting third-party AI agents accessing customer accounts, including instances where agents appeared to attempt actions beyond their users’ instructions.
Morgan Stanley has blocked third-party agents but has not identified any unauthorised or rogue activity, according to a person familiar with the bank. Charles Schwab has also been blocking large external agents, with chief executive Rick Wurster saying the brokerage prevents dozens of attempts each week.
Schwab is exploring a more controlled approach that would allow external AI models to retrieve account balances and conduct research through its own Model Context Protocol (MCP) servers. However, these integrations would not be permitted to transfer funds or execute transactions.
Banks Warn Customers About Credential Sharing
JPMorgan Chase has made clear that it considers sharing banking credentials with third-party services insecure. During its third-quarter earnings call, chief financial officer Jeremy Barnum said the bank blocks these services where possible and that customers could be held responsible for activity resulting from giving agents their login details.
Chief executive Jamie Dimon said the bank is not opposed to AI agents themselves, noting that JPMorgan already uses them internally and intends to introduce customer-facing applications.
Bank of America has issued similar warnings, telling customers that they remain responsible for activity resulting from shared credentials.
Wells Fargo has also updated its Online Access Agreement to address AI agents. Under the revised terms, instructions and transactions performed by an agent using a customer’s login credentials or granted authority may be treated as the customer’s own actions, subject to applicable law. Customers may be held responsible for unintended transactions and actions that exceed the authority they intended to provide.
Open Banking Rules Remain Stalled
The dispute over AI agent access is closely connected to a longer-running disagreement between banks and financial technology companies over how customer account information should be shared.
For years, many budgeting and payment applications have relied on screen scraping, a process that involves using customer-provided credentials to access online banking portals and collect account information.
The Consumer Financial Protection Bureau’s Section 1033 rule was designed to move the industry away from this practice. Finalised in October 2024, the regulation requires financial institutions to establish developer interfaces, typically APIs, through which authorised third parties can securely access customer financial data.
The rule also restricts banks from relying on credential-based screen scraping to satisfy their data-sharing obligations.
However, implementation has been delayed. A federal court in Kentucky blocked the CFPB from enforcing the rule, while the Bureau began reconsidering its requirements. A revised proposal has been submitted for White House regulatory review but has not yet been published.
With the regulation stalled, financial institutions have greater discretion over how third-party AI agents interact with their systems.
Industry Develops Standards for AI Agent Access
While regulatory uncertainty continues, financial institutions and technology companies are exploring alternative approaches to managing AI agent permissions.
Bank of America, BBVA, Synchrony and Wells Fargo have joined the Personal Agent Protocol, an open standard led by Sierra and Meta. The initiative aims to establish a consistent method for AI agents to identify themselves to businesses, verify whom they represent and operate within permissions granted by customers.
Built on OAuth, the protocol allows customers to determine whether agents receive read-only access or permission to perform transactions.
BBVA has highlighted the limitations of existing approaches, noting that many AI agents currently need to navigate websites designed for human users or rely on customers’ credentials to complete tasks.
The initiative includes 35 design partners spanning technology, financial services and e-commerce, reflecting growing interest in establishing common standards for agent-driven interactions.
Security Concerns or Financial Incentives?
The restrictions have also raised questions about whether banks are primarily concerned with protecting customers or preserving their own financial interests.
Jim Bianco, president of Bianco Research, has argued that AI agents could make it easier for customers to move idle deposits into higher-yielding alternatives such as money market funds. This could threaten an important source of revenue for banks that benefit from holding customer deposits at relatively low interest rates.
JPMorgan’s Dimon has defended the restrictions on security grounds, arguing that allowing external agents into banking systems creates unacceptable risks for both customers and financial institutions.
The distinction between protecting account security and protecting traditional banking revenue is likely to remain a point of debate as AI agents become more widely adopted.
The Future of AI-Powered Financial Services
The growing use of AI agents in financial services is creating pressure for more consistent approaches to authentication, permissions and transaction oversight.
Until the Section 1033 framework is revised or industry standards become more widely adopted, companies developing AI-powered financial tools will continue to face different access requirements across individual banks.
The challenge will be establishing a system that allows customers to delegate financial tasks to AI agents without compromising account security, control over transactions or the ability to revoke access.
For banks, technology providers and regulators, the next stage of AI adoption in financial services will depend on finding a balance between innovation, customer autonomy and financial security.
Source: https://www.artificialintelligence-news.com/news/banks-block-ai-agents-open-banking-rule/
