The Loaded Wheel: Why the First AI Crisis Will Be Governed by Finance, and What That Does to AI Governance

Abstract

The argument that only a crisis will break the deadlock in AI governance treats the crisis as a roulette wheel: exogenous in kind, unknowable in advance. I argue the wheel is loaded — not in its timing, which is genuinely beyond reach, but in its governor. Scored on the focusing-event properties that determine whether a candidate crisis opens a policy window — legibility of harm, speed of attribution, and maturity of the receiving regulator — finance wins the combination, decisively on the third axis. Whatever domain the first legible AI-involved systemic event originates in, the modal outcome is that it routes to financial regulators, who produce governance faster than any other receiver could — and mis-scoped: sectoral, keyed to the legible failure, and structurally unable to reach the model layer the AI-governance community wants governed. I derive design principles for the pre-crisis shelf and price four borrowings from financial regulatory machinery against their own pathologies.