In a recent commentary, Torsten Slok, a senior economist at Apollo Global Management, sounded the alarm about a new, technology‑driven risk to the stability of the banking system. He argued that the rapid rise of artificial‑intelligence‑powered financial agents could create a novel pathway for a classic bank‑run scenario, but one that would be triggered automatically, without the need for panic‑driven human behavior.
Traditionally, a bank run occurs when a large number of depositors, fearing that a bank might become insolvent, rush to withdraw their money at the same time. The sudden outflow can deplete a bank’s liquid reserves, forcing it to sell assets at fire‑sale prices, and potentially leading to a cascade of failures across the financial sector.
Slok’s concern is that AI agents—software programs that can execute financial transactions on behalf of users—could automate this process, turning a gradual shift in preferences into an abrupt, system‑wide shock. The mechanism he describes is straightforward yet unsettling. Many households keep a portion of their cash in checking accounts that pay little or no interest.
These accounts are convenient for everyday spending but are not optimized for maximizing returns. As AI‑driven personal finance tools become more sophisticated, they are increasingly capable of scanning the market for better‑yielding alternatives, such as high‑interest savings accounts, money‑market funds, short‑term Treasury bills, or even decentralized finance (DeFi) protocols offering attractive rates. When the AI detects a spread that exceeds a certain threshold, it can automatically initiate a transfer of funds from the low‑interest checking account to the higher‑yield option. If a significant share of the population adopts such agents, the cumulative effect could be massive.
Imagine that 30 % of checking‑account balances in the United States—amounting to trillions of dollars—are managed by AI tools programmed to chase the best rates. When a new, higher‑yield product is launched, these agents could simultaneously move large volumes of cash, draining liquidity from traditional banks in a matter of minutes or hours.
Unlike a conventional bank run, which is often spurred by rumors or a loss of confidence, this AI‑driven outflow would be triggered by a purely rational, algorithmic decision: the pursuit of higher returns. The speed and scale of such transfers present several challenges for banks and regulators. First, banks rely on a stable base of low‑interest deposits to fund loans and other assets. A sudden, algorithm‑driven exodus would force them to liquidate assets quickly, potentially at a loss, thereby eroding capital buffers.
Second, the predictability of the triggers—such as a new rate announcement from a competitor—could allow market participants to anticipate the outflows and position themselves advantageously, further amplifying the impact. Third, the automated nature of the transfers reduces the window for banks to intervene or communicate with customers, making traditional crisis‑management tools less effective. Regulators have already begun to consider the implications of AI in finance, focusing largely on issues like algorithmic bias, transparency, and systemic risk from high‑frequency trading. Slok’s warning expands that conversation to include the risk of automated deposit flight.
Existing safeguards, such as reserve requirements and deposit insurance, may need to be re‑examined in light of this new dynamic. For example, deposit insurance limits are designed to protect individual depositors up to a certain amount, but they do not address the collective impact of millions of small, automated withdrawals occurring simultaneously. Potential mitigations could involve a combination of policy and technological solutions. On the policy side, central banks might consider adjusting reserve ratios or introducing liquidity buffers specifically earmarked for rapid outflows.
They could also explore new forms of insurance or guarantees that cover not just individual depositors but also the systemic risk posed by mass automated transfers. On the technology side, banks could develop their own AI agents that monitor deposit flows in real time and provide early warnings of abnormal patterns.
Moreover, they could offer competitive, AI‑compatible products that retain deposits by matching the speed and convenience of third‑party agents. Another angle is consumer education. While AI agents promise convenience, users should be made aware of the broader implications of moving large sums of money automatically. Financial literacy programs could emphasize the importance of diversification—not just across asset classes, but also across institutions—to reduce concentration risk.
In addition to the immediate liquidity concerns, there are longer‑term macroeconomic implications. If AI agents systematically redirect deposits toward higher‑yielding, often less regulated venues, the traditional banking sector could see a gradual erosion of its funding base. This shift might encourage banks to pursue riskier lending strategies to maintain profitability, potentially sowing the seeds for future instability. Conversely, if banks adapt by offering more attractive rates and digital services, the competition could lead to a more efficient allocation of capital, benefiting consumers overall.
Slok’s cautionary note underscores the need for a proactive approach. The financial industry has historically been adept at responding to technological disruption—consider the evolution from paper checks to electronic transfers, or the rise of online banking.
However, the speed at which AI can act, coupled with its ability to operate across vast numbers of accounts simultaneously, represents a qualitatively different challenge. It calls for coordinated action among banks, fintech firms, regulators, and even the developers of the AI agents themselves. In conclusion, while AI agents hold great promise for optimizing personal finance and enhancing consumer convenience, they also introduce a novel vector for systemic risk. By automatically reallocating cash from low‑interest checking accounts to higher‑yield alternatives, these agents could, in theory, trigger a rapid, large‑scale withdrawal of deposits that mirrors a traditional bank run but unfolds at the speed of algorithms.
Torsten Slok’s warning serves as an early alert that the industry must heed, prompting a reassessment of liquidity management, regulatory frameworks, and consumer education to ensure that the benefits of AI do not come at the expense of financial stability.