Wall Street firm warns AI agents could trigger bank run

A major Wall Street firm warned in a recent research note that autonomous AI agents could trigger a new kind of bank run by moving large volumes of deposits and trading flows over minutes or hours.

AI agents are software that act on behalf of users to manage money and execute transactions. The research described scenarios in which widely used agents, programmed to optimize returns or cash access for millions of users, detect a risk signal and quickly shift funds from bank accounts into other assets or platforms.

Analysts modeled cases where multiple popular consumer agents reacted simultaneously to the same market signal, such as a sharp drop in commercial paper yields or a perceived credit event. In those simulations, deposits and money market balances moved to perceived safer places or into cryptocurrencies and brokerage accounts within a compressed period. The note said that pace could strain intraday settlement and force banks to draw emergency liquidity or curtail lending to meet outflows.

The potential stress is not limited to retail deposits. Agents operating on institutional platforms could accelerate runs on money market funds, prime cash funds and short-term funding conduits that support commercial paper and other corporate financing. The research highlighted the speed and correlation of algorithmic decision-making as the key difference from past deposit runs: identical rules and data can produce highly correlated flows.

The note recommended specific steps for banks, fund managers and regulators, including upgraded real-time monitoring of flow concentrations tied to third-party agents, rate limits or throttles on automated transfers, stronger authentication for machine-initiated transactions, and stress-testing exercises that include rapid, correlated withdrawals by connected software. It urged coordinated industry work to set standards for how AI agents access customer accounts and for signals that might trigger mass reallocations.

The research warned: “If large numbers of autonomous agents are given leeway to move funds in response to the same triggers, the resulting speed and correlation of flows could exceed the capacity of current liquidity frameworks to respond.” It said trusted guardrails and clearer accountability for firms that deploy or host agents could reduce the risk of unintended cascades.

Use of AI assistants for finance has expanded as tech companies and fintechs build agents that can shop for investment options, move cash between accounts for yield or automatically rebalance portfolios. The firm called for regulators to include AI-driven outflows in liquidity stress tests and for industry groups to develop norms such as maximum transfer sizes and pause mechanisms when market stress is detected.

The note also raised legal and operational questions about where responsibility will sit when an autonomous agent acting for a customer causes a rapid redeployment of funds, and called for clarity on accountability among banks, fintech platforms and AI providers to design effective mitigants.