The chief economist at Apollo Global Management warns that agentic AI could trigger a bank run. The prediction comes from Torsten Slok in a brief research note. The implications extend far beyond the technology sector.
Agentic AI refers to systems that act independently on behalf of users. These agents can execute financial transactions without direct human oversight. Their speed and autonomy set them apart from conventional software.
A bank run occurs when depositors rush to withdraw funds simultaneously. Traditional runs unfold over days or weeks as crowds gather at branches. AI agents could compress that timeline into minutes.
Slok’s concern centers on coordinated behavior among autonomous systems. Thousands of AI agents might react to the same market signal at once. Their collective withdrawals could drain a bank before regulators respond.
The risk mirrors earlier episodes of algorithmic trading. Flash crashes in equity markets showed how automated systems amplify volatility. Agentic AI introduces similar dangers to retail and commercial banking.
Current regulations do not address machine-initiated withdrawals directly. Deposit insurance protects individual accounts but may not handle mass simultaneous exits. Supervisors have few tools to slow down digital runs.
Banks are already deploying AI for customer service and fraud detection. Extending those systems to transaction decisions raises new stability questions. Slok’s note suggests the industry needs to prepare now.
The warning is not a prediction of imminent collapse. It highlights a structural vulnerability that deserves attention. Policymakers and bank executives should treat it as a planning scenario.
