In a recent commentary, Torsten Slok, a senior economist at Apollo Global Management, raised a flag about a new, technology‑driven risk to the stability of the banking system. He argued that the rise of sophisticated artificial‑intelligence agents—software programs that can act autonomously on behalf of consumers—could create a scenario that resembles a classic bank run, but one that is triggered not by panic or rumor, but by algorithmic decision‑making. The core of Slok’s warning centers on the fact that many households keep a substantial portion of their liquid wealth in checking or other low‑interest deposit accounts.
These accounts are attractive because they offer immediate access to cash, but they typically pay little or no interest, especially in a low‑rate environment. At the same time, the financial market has seen a proliferation of higher‑yield alternatives, ranging from online savings accounts and money‑market funds to short‑term bond ETFs and even tokenized assets that promise returns well above the traditional bank rates.
AI agents, which are increasingly embedded in personal finance apps, robo‑advisors, and even voice‑activated assistants, are designed to optimize the financial outcomes for their users. They constantly scan the market for better yields, compare fees, and execute transactions with minimal human oversight. When an AI agent detects a gap between the interest earned on a checking balance and the higher returns available elsewhere, it can automatically transfer the funds, often in fractions of a second. If millions of such agents operate under similar optimization rules, the cumulative effect could be dramatic.
Imagine a scenario where a new, higher‑yield product is announced. Within minutes, AI agents across the country could begin moving money out of traditional checking accounts and into the new vehicle.
The speed and scale of these transfers would be far greater than any historical bank‑run episode, which relied on people physically lining up at a branch or making phone calls. Instead, the outflow would be executed electronically, overwhelming the liquidity management systems of banks that are accustomed to more gradual withdrawals. Slok points out that the traditional safeguards against runs—such as deposit insurance, liquidity buffers, and central‑bank backstops—might be less effective in this new context. Deposit insurance protects individual savers, but it does not stop the rapid, collective reallocation of funds that an AI‑driven system could initiate.
Liquidity buffers are designed to cover a certain percentage of expected withdrawals based on historical patterns; they may be insufficient when the withdrawal pattern is dictated by algorithmic triggers rather than human behavior. Moreover, central banks may find it challenging to intervene quickly enough to stem a cascade that unfolds in milliseconds.
The implications extend beyond the immediate liquidity strain. Banks rely on a stable base of low‑cost deposits to fund longer‑term lending activities, such as mortgages and business loans.
A sudden erosion of that deposit base could force banks to raise the cost of borrowing, tighten credit standards, or even curtail lending altogether. This, in turn, could have a chilling effect on economic growth, particularly if the phenomenon coincides with a broader tightening of monetary policy. Regulators are beginning to take notice.
Some central banks and supervisory agencies have started to explore how to monitor algorithmic behavior in the financial sector. They are considering requirements for transparency—such as mandatory disclosures about the criteria that AI agents use to move funds—and for stress‑testing that incorporates scenarios where large volumes of deposits are withdrawn automatically. However, crafting rules that address the speed and opacity of AI decision‑making without stifling innovation remains a delicate balancing act.
From a consumer perspective, the allure of higher yields is understandable, especially when traditional bank rates have lingered near zero for years. Yet, the convenience of letting an AI handle the optimization may obscure the systemic risks. Users may not realize that their personal financial decisions, when aggregated across millions of accounts, can collectively influence the stability of the banking system.
Education and awareness campaigns could help individuals understand the broader impact of their automated choices. In addition to the direct risk of a liquidity crunch, there are secondary concerns. Rapid fund movements could exacerbate market volatility, as large inflows into a particular asset class push prices up quickly, potentially creating bubbles.
Conversely, sudden outflows could depress asset prices, triggering losses for investors who are not able to react as swiftly as the AI agents. The feedback loop between banking stability and market dynamics could become more pronounced, making the financial system more interconnected and, paradoxically, more fragile.
To mitigate these risks, Slok suggests several possible approaches. First, banks could develop their own AI‑driven tools that offer competitive yields, thereby retaining deposits within the institution while still meeting the optimization goals of customers.
Second, a coordinated industry effort could establish “cool‑down” periods for large automated transfers, similar to settlement windows used in securities markets, giving banks time to manage liquidity. Third, policymakers could consider revising the definition of a deposit run to explicitly include algorithmic triggers, ensuring that emergency liquidity facilities are designed to respond to this new form of stress.
In conclusion, while AI agents promise greater efficiency and better financial outcomes for individual consumers, they also introduce a novel systemic vulnerability. Torsten Slok’s warning serves as a reminder that technological progress must be accompanied by thoughtful regulation and proactive risk management. By recognizing the potential for AI‑driven deposit outflows and implementing safeguards now, the financial industry can harness the benefits of automation without compromising the stability that underpins the broader economy.