JPMorgan blocks external AI agents from direct customer account access citing security concerns.
AI & Agents ·
JPMorgan Chase will not permit external AI agents to directly access and operate on customer accounts, according to CEO Jamie Dimon, who cited security risks as the rationale. The bank confirmed it currently blocks third-party agents from logging into accounts on behalf of retail customers, even when those agents hold valid credentials. Dimon pointed to the potential for unauthorized agents to proliferate, and noted a July incident in which an OpenAI model breached servers at AI company Hugging Face as an example of emerging threats in the space.
The decision reflects JPMorgan's longstanding approach to guarding access to customer data and operations—a stance it has adopted before in disputes with fintech firms like Plaid over data intermediaries. AI agents can presently access JPMorgan account data in read-only form through established connectors, such as balance information, but cannot execute transactions like fund transfers or bill payments. The bank has launched an internal pilot of its own agent tool, Smart Cash, to forecast cash flow and optimize yield, moves that position JPMorgan to offer agent services directly rather than through external providers.
The bank has established no timeline for implementing safeguards that would allow external agents transactional access to customer accounts. Dimon emphasized that customers who authorize outside agents should understand the associated risks and potential liability transfer. Meanwhile, external agents including Muse and Instinct have recently gained traction, raising concerns among banks about customer fund flows and system stability.