AI agents, bank runs, and “kill billions”: what the new AI risk wave means for markets

On September 28, 2026, Apollo’s Torsten Slok warned that AI agents could accelerate a bank-run dynamic by automatically moving household cash from low-yield checking accounts into higher-yield alternatives. The mechanism is straightforward but destabilizing: if AI systems optimize returns in real time, depositors may react faster than banks can manage liquidity. In parallel, Bill Gates cautioned—via a report carried by The Jerusalem Post—that AI could be weaponized by bad actors with catastrophic consequences, using language that underscores extreme tail risk. A separate item described AI users spanning a conservative Christian in Texas, a witchcraft practitioner in Virginia, and a Jewish man from California, highlighting how broad and socially diverse AI adoption is becoming. Geopolitically, the cluster points to a shift from “AI as productivity” toward “AI as an accelerant of systemic risk.” Slok’s warning frames AI as a financial stability threat, not by changing credit fundamentals but by changing the speed and coordination of consumer behavior. Gates’ warning elevates the security dimension, implying that AI can lower barriers for malicious planning, persuasion, and operational execution, which in turn raises the stakes for governments and critical infrastructure operators. The J.P. Morgan note that global AI trade could revive after a recent pullback adds a counterweight: capital markets still expect growth in AI supply chains, but risk premiums may rise if security and financial-stability concerns intensify. Overall, the winners are likely firms with strong compliance, monitoring, and liquidity management, while losers include deposit-heavy institutions and jurisdictions that lag in AI governance. Market implications cut across banking, cybersecurity, and AI supply chains. If AI-driven deposit switching becomes credible, it can pressure bank funding costs and widen spreads for institutions perceived as vulnerable to rapid outflows, potentially lifting money-market volatility and increasing demand for higher-yield instruments. On the AI trade side, J.P. Morgan’s view suggests a rebound in cross-border AI hardware and services flows, which typically supports semiconductors, data-center construction, networking equipment, and cloud infrastructure demand. The “kill billions” framing also tends to boost hedging behavior—raising demand for cyber insurance, security tooling, and risk-management services—while increasing volatility in AI-adjacent equities during policy or incident headlines. While the articles do not quantify price moves directly, the direction is clear: higher perceived tail risk can increase discount rates for unmitigated AI deployments and raise near-term risk premia. What to watch next is whether regulators and financial supervisors treat AI-driven deposit mobility as a liquidity-stability issue rather than a consumer-behavior curiosity. Key indicators include evidence of faster-than-usual deposit reallocation, changes in bank funding spreads, and any guidance on AI agent governance for consumer finance. On the security front, monitor for concrete policy actions—such as model access controls, provenance requirements, and incident reporting—after high-profile warnings like Gates’ and after any public demonstrations of misuse. For markets, track J.P. Morgan’s “revive” thesis through trade data, export controls, and order-flow signals in AI hardware and cloud capacity, looking for confirmation or reversal after the recent pullback. The escalation trigger would be a real-world incident that links AI agents to measurable financial instability or large-scale harm, while de-escalation would come from credible mitigations, audits, and rapid containment after any misuse reports.

Geopolitical Implications

• 01

  AI governance is becoming a cross-border strategic issue: security controls and financial-stability rules may diverge by jurisdiction, affecting trade and investment flows.

• 02

  Faster AI-enabled consumer and operational actions can amplify systemic risk, increasing pressure on central banks and supervisors to treat AI as a stability variable.

• 03

  Security-focused narratives around AI misuse can accelerate export-control and model-access regimes, reshaping the global AI supply chain.

Key Signals

• —Evidence of faster deposit reallocation and widening bank funding spreads tied to AI-enabled financial automation.
• —Regulatory announcements on model access, provenance, and incident reporting for high-risk AI capabilities.
• —Trade and order-flow indicators for AI hardware/services that confirm or contradict J.P. Morgan’s revival thesis.
• —Cybersecurity demand spikes (insurance pricing, security tooling procurement) following misuse headlines.