Apollo economist warns AI agents could trigger bank runs

By Splitfeed EditorialLast updated Sep 29, 2026, 13:51 UTC

On September 27, 2026, Apollo Global Management chief economist Torsten Slok issued a warning that AI agents could trigger what some have taken to calling an "agentic bank run." The scenario he described involves AI agents, acting autonomously on behalf of users, scanning available options and moving customer deposits out of low-yield accounts, some paying as little as 0.1 percent, into higher-paying alternatives offering rates around 5 percent. One outlet pointed to Meta's Muse as an example of the kind of agent capable of this behavior. To frame the scale of risk, a 24/7 Wall St. report referenced the fastest bank run in U.S. history, in which customers pulled $42 billion in deposits in a single day.

The concern cuts to something structural in how banks operate. The low-yield deposits that Slok is worried about losing are what banks call cheap deposits, a low-cost funding base that makes lending affordable and profitable. If AI agents optimize this away at machine speed, the effect on bank funding and lending costs could be significant, and the speed involved removes the friction that historically gave institutions time to respond. What remains genuinely in dispute is whether this is an imminent threat or a speculative scenario, and whether regulators or banks have tools adequate to the pace of a machine-driven outflow. No regulator or bank has publicly responded to Slok's warning.

The Anti-AI camp has engaged with this story most directly. Accounts including politikrypt.bsky.social described AI agents automating the withdrawal of cheap deposits and potentially triggering a systemic liquidity crisis. CryptoSnake framed it as AI systems optimizing toward higher yields in ways that could raise bank funding and lending costs. Financier.News put specific numbers on it, citing the movement of cash from 0.1 percent bank accounts into 5 percent yields. The Middle Ground position, reflected by thoughtsnprayers.bsky.social, acknowledges the mechanism Slok describes but treats it as a forward-looking risk rather than a present crisis, noting that the problem would materialize only if the migration happened fast. The Pro-AI camp has not reacted to this story; typically, that camp would be expected to argue that AI-driven financial optimization benefits consumers by ensuring they earn competitive returns on their own money.

The argument would sharpen considerably if a major regulator, the Federal Reserve, the FDIC, or the OCC, were to comment on agentic deposit behavior, or if a bank disclosed material deposit outflows linked to AI-managed accounts. Any formal rulemaking around AI agent permissions in financial services would also move the debate from speculation toward policy.