Major US financial institutions, including Morgan Stanley, JPMorgan, Bank of America, and Wells Fargo, are actively blocking external autonomous AI agents that attempt to log into customer accounts using user-provided credentials. According to the institutions, this practice violates core security policies that require account holders to authenticate themselves personally. As these systems increasingly attempt to operate without direct, real-time human oversight, the friction between automated software and legacy security perimeters has escalated sharply. Institutions like Morgan Stanley have moved to block such access entirely, while peers including JPMorgan, Bank of America, and Citi follow suit to prevent unauthorized session handling.
At the same time, as Morgan Stanley noted, systems have not yet recorded widespread malicious activity from autonomous software on their platforms, yet the architectural risk remains unacceptable. The current infrastructure assumes a human sits behind every session token, leaving compliance officers blind to whether a transaction is driven by intent or an algorithmic loop.
Read-Only Compromises
To bridge the gap, major lenders are offering a narrow concession: read-only access that permits agents to scrape and aggregate data while strictly barring any transactional capabilities. Financial institutions explain that while customer demand for automated financial assistants is clear, the risk profile of unmonitored execution leaves no room for error. Most compliance teams admit they cannot reliably distinguish between an agent executing a user's pre-configured workflow and an autonomous script operating outside its parameters.
Most banks want to limit agents to read-only access, letting them view account data but not carry out any actions.
Workarounds persist as users route agents through browser-automation tools, such as computer-use interfaces or open-source frameworks like OpenClaw, effectively mimicking human input at the DOM level. Bypassing bank-side API controls through visual browser automation exposes an architectural blind spot that traditional firewalls were never designed to catch.
Market Integration Realities
The standoff highlights a structural deficit in how software vendors approach enterprise security boundaries. Rather than waiting for API standardization, developers of autonomous software are opting to force user-side authentication bypasses. In response to these barriers, infrastructure providers are pivoting; OpenAI, for instance, has launched a dedicated ChatGPT service tailored for financial workflows in the US, deliberately restricted to read-only financial data access to comply with institutional walls.
Financial institutions are drawing a hard line at the transaction boundary, forcing AI developers to accept a severely throttled feature set or build entirely within sanctioned, API-driven sandboxes. For enterprise software buyers, this signals that fully autonomous financial agents will remain stalled until banks establish regulated, machine-to-machine payment rails.
Editorial analysis
### We covered this earlier
On 22 October 2026, automated purchases required manual confirmation by default, with users able to pre-approve transactions under a set cap such as $100. Major financial institutions including Morgan Stanley, JPMorgan, Bank of America, and Wells Fargo are now actively blocking third-party autonomous AI agents that attempt to access customer accounts using provided credentials. The banks state that these systems operating without direct human control violate security rules requiring account owners to authenticate in person.
Our stories: 22.10.2026
Compiled by AI from our stories, other outlets and the primary source; each line shows its source