In a recent commentary, Torsten Slok, chief economist at Apollo Global Management, highlighted a growing concern that advanced artificial‑intelligence agents could inadvertently set off a modern‑day bank run. The core of his warning revolves around the ability of AI‑driven platforms to monitor and act on the most favorable financial products in real time, moving consumer cash from traditional, low‑interest checking accounts into alternatives that promise higher yields.
While the technology promises greater efficiency and better returns for savers, it also introduces systemic risks that regulators and financial institutions have not yet fully addressed. Slok explains that the traditional banking model relies on a stable base of deposits—funds that remain relatively static over short periods, providing banks with the liquidity needed to fund loans and other assets.
Historically, the inertia of human decision‑making—people checking balances once a month, reacting to interest‑rate changes only after a noticeable lag—has acted as a natural buffer against sudden, large‑scale withdrawals. However, AI agents are designed to eliminate that inertia. By continuously scanning the market for the best rates, they can automatically trigger transfers the moment a more attractive option appears, whether that be a high‑yield savings account, a money‑market fund, or a short‑term bond.
The implications are profound. Imagine millions of households each holding a modest sum in a checking account that earns a meager 0.1 percent interest. An AI assistant, programmed to maximize net returns, would detect that a new online bank offers 4.5 percent APY on a comparable product.
Within seconds, the algorithm could initiate a transfer, moving the funds to the higher‑yield account. Multiply that action across a large portion of the population, and the result is a rapid, coordinated outflow of cash from traditional banks.
In the worst‑case scenario, the speed and scale of these withdrawals could outpace a bank’s ability to meet demand, forcing it to sell assets at a loss or seek emergency liquidity. Slok points out that this risk is not merely theoretical. The financial sector has already witnessed the power of algorithmic trading to create flash crashes in equity markets, where automated systems amplify price movements within milliseconds. The same principle can apply to retail banking deposits.
The difference is that deposits are a core funding source for banks, and a sudden shortfall can compromise their capacity to lend, potentially tightening credit conditions for businesses and consumers alike. Regulators are beginning to take note. The Federal Reserve and other supervisory bodies have started to explore how AI‑driven financial advice and robo‑advisors fit within existing frameworks.
Yet, the focus has largely been on consumer protection—ensuring that advice is suitable and that data privacy is maintained—rather than on systemic stability. Slok argues that a more holistic approach is needed, one that considers the aggregate impact of millions of small, automated decisions. Potential mitigation strategies include: 1.
**Liquidity Buffers:** Banks could be required to hold larger high‑quality liquid assets to absorb sudden deposit outflows without distress. 2. **Rate Caps on Automated Transfers:** Financial platforms might implement limits on how frequently an AI can move funds without explicit user confirmation, introducing a friction that slows the cascade. 3.
**Transparency Requirements:** Mandating that AI agents disclose the criteria they use for moving money could help users understand the risk and make more informed choices. 4. **Co‑ordination with Central Banks:** In the event of a rapid shift, central banks could provide temporary liquidity facilities to banks experiencing unexpected withdrawals. Beyond regulatory measures, there is a cultural component.
Consumers need to be educated about the trade‑off between higher yields and the stability that comes from keeping funds within a traditional banking relationship. While a 4‑5 percent return is attractive, it often comes with different risk profiles, such as limited FDIC coverage or exposure to market volatility. AI agents, by design, prioritize numerical optimization and may not fully account for these nuances unless explicitly programmed to do so.
Slok also emphasizes that the technology itself is not inherently dangerous; rather, it is the speed and scale at which it can operate that creates new challenges. He suggests that banks could turn this to their advantage by offering competitive digital products that integrate AI‑driven personalization while maintaining deposit stability. For example, banks could develop hybrid accounts that automatically allocate a portion of a customer’s balance to higher‑yield instruments within the same institution, reducing the incentive for external transfers.
In summary, the rise of AI agents capable of autonomously reallocating household cash poses a novel threat to the traditional deposit base that underpins the banking system. By moving money at unprecedented speed from low‑interest checking accounts to higher‑yield alternatives, these agents could inadvertently trigger a cascade of withdrawals reminiscent of a classic bank run, but executed in milliseconds rather than days. Addressing this risk will require a combination of regulatory foresight, enhanced liquidity safeguards, and consumer education to ensure that the benefits of AI‑enhanced financial management do not come at the expense of systemic stability.