In a recent commentary, Torsten Slok, chief economist at Apollo Global Management, sounded the alarm on a potential new risk to the banking sector: the rise of autonomous artificial‑intelligence agents that could, without human prompting, relocate vast sums of cash from traditional, low‑interest deposit accounts to more lucrative investment vehicles. While the concept may sound futuristic, the underlying mechanics are already in place, and the implications could be profound if these digital actors begin to act en masse.
At the heart of Slok’s warning is the observation that many households keep a substantial portion of their liquid wealth in checking or savings accounts that earn virtually no interest. In a low‑rate environment, these accounts function more as a convenient store of value than as a source of income. Historically, the decision to move money out of such accounts has been driven by human judgment—people compare rates, read financial news, and decide whether to open a higher‑yield certificate of deposit, a money‑market fund, or a short‑term bond. However, the advent of sophisticated AI algorithms that can continuously monitor interest‑rate differentials, fees, and market conditions changes the calculus dramatically.
Imagine a scenario in which an AI‑powered personal finance assistant is linked to a user’s bank account. The assistant constantly scans the financial landscape, noting that a new online‑only bank is offering a 5% annual percentage yield on a savings product, while the user’s current checking account yields 0.01%. Within seconds, the AI can generate a recommendation, draft the necessary paperwork, and even initiate the transfer, all without the user needing to lift a finger.
Multiply this capability across millions of users, and the speed at which deposits could be pulled from traditional banks becomes staggering. Slok points out that the danger is not merely theoretical.
In the past, bank runs have been triggered by panic, rumors, or sudden loss of confidence. The classic 1930s episodes in the United States, for instance, saw depositors line up at teller windows, withdrawing cash en masse and forcing banks into insolvency. Modern banking regulations and deposit insurance have largely mitigated the risk of such panics, but the underlying vulnerability—reliance on a stable base of low‑interest deposits—remains. AI agents could effectively automate the decision to move money, bypassing the emotional hesitancy that often slows human‑driven runs.
The result could be a rapid, algorithmic outflow that leaves banks scrambling to replace lost funding. One of the key mechanisms that could amplify this effect is the integration of AI with open‑banking APIs. In many jurisdictions, regulators have mandated that banks provide secure, standardized interfaces that allow third‑party applications to access account information and initiate transactions with consumer consent. While these APIs were designed to foster competition and innovation, they also create a conduit through which AI agents can execute transfers at scale.
A single well‑written script could, in theory, trigger thousands of transfers in a matter of minutes. The potential consequences for banks are multifaceted. First, a sudden drop in deposits would force institutions to seek alternative funding sources, often at higher costs. They might need to tap the interbank market, issue short‑term debt, or sell assets at unfavorable prices.
Second, the loss of cheap funding could erode profit margins, prompting banks to raise fees on remaining customers or to tighten lending standards, which could have broader macroeconomic repercussions. Third, the perception of instability—whether justified or not—could lead to a feedback loop, where other depositors, observing the outflows, decide to withdraw their own funds, further intensifying the strain. To mitigate these risks, Slok suggests several proactive measures. Regulators could consider imposing limits on the frequency or volume of automated transfers, especially those initiated by third‑party services.
Banks might enhance their own AI capabilities to detect unusual patterns of outflows and respond in real time, perhaps by offering instant, higher‑yield alternatives to retain customers. Additionally, greater transparency around the terms and conditions of high‑yield products could help consumers make more informed choices, reducing the allure of “too‑good‑to‑be‑true” offers that may be promoted by aggressive AI agents. From a broader perspective, the issue underscores a fundamental tension in the digital age: the same technologies that empower consumers with better financial tools also introduce new systemic vulnerabilities.
As AI becomes more embedded in everyday financial decisions, the line between personal choice and algorithmic suggestion blurs. Consumers may unwittingly become participants in a collective behavior driven by code rather than conscious deliberation. In conclusion, while AI agents hold the promise of greater efficiency and personalized financial management, they also present a novel challenge to banking stability.
Torsten Slok’s warning serves as a reminder that policymakers, financial institutions, and technology developers must work together to design safeguards that prevent rapid, automated deposit withdrawals from turning into a modern‑day bank run. By anticipating these dynamics and implementing thoughtful controls, the industry can harness the benefits of AI without compromising the foundational trust that underpins the banking system.