Artificial intelligence (AI) agents could prompt a new form of bank run by enabling consumers to automatically transfer funds from low-yield checking accounts to higher-interest alternatives, according to Torsten Sløk, chief economist at Apollo Global Management. Sløk issued the warning in a note published on Sunday, highlighting the growing influence of AI tools that manage personal finances with increasing autonomy.

This emerging risk stems from personal AI agents, such as Meta’s recently launched Muse platform, which have access to users’ financial data and can perform complex tasks on their behalf beyond answering queries. By optimizing cash management, these AI agents may encourage widespread shifts of deposits from traditional bank accounts offering minimal interest to higher-yield fintech and online savings products.

Sløk emphasized that banks rely heavily on cheap deposits from customers’ checking and savings accounts to fund loans. When deposits remain in low-paying accounts, banks can pay depositors limited interest while lending at higher rates, sustaining profitability. However, if AI agents facilitate rapid and large-scale withdrawal of these funds to platforms offering better returns, this business model could be undermined, potentially posing risks to the broader financial system.

The economist illustrated the incentive for consumers with an example: a $10,000 balance earning 0.1% interest in a checking account yields roughly $10 annually, but at a 5% return, the same amount generates about $500. His note identified 11 digital financial services offering rates ranging from 3.3% to 5%, with AdelFi providing the highest rate at 5% and SoFi at 4.5%. These figures sharply contrast with national FDIC averages, where the savings account rate stands at 0.4% and checking accounts at only 0.1%.

Meta introduced Muse on September 8, promoting its capacity to execute a wide range of actions for users, moving beyond standard AI chatbots. Financial data aggregator Plaid confirmed that it supports Muse by supplying access to user-approved information from over 12,000 U.S. financial institutions and applications, including account balances, transaction histories, investment portfolios, and mortgage information.

While Meta did not comment on the potential impact of AI agents on banking stability, the development underscores an evolving dynamic where AI-driven financial decision-making may alter consumer behavior and liquidity within traditional banking systems. As AI tools increasingly manage routine financial matters, their influence on deposit patterns and bank funding models will warrant close attention from regulators and industry stakeholders.