Compliance Moral Hazard and the Backfiring Mandate

The buried lede is the 'compliance moral hazard' framing: institutions in an AML network have rational incentives to free-ride on others' reporting effort, since flagging your own customers carries regulatory and reputational cost while staying quiet lets competitors absorb that burden. The TVA mechanism is specifically designed to make honest reporting the dominant strategy even when institutions know others are watching.

This connects most directly to the CoopEval benchmark covered on April 16, which tested whether agents in public goods games and prisoner's dilemma scenarios could sustain cooperation under game-theoretic mechanisms. That paper found LLM agents consistently defect; this AML paper is essentially asking the same question about human institutions, and arriving at a mechanism-design answer rather than a behavioral one. The difference matters: where CoopEval evaluated whether cooperation could be nudged, this work tries to make defection structurally irrational through incentive-compatible scoring. Both papers are working on the same underlying problem, which is that decentralized agents with private information tend to withhold it when sharing is costly.

Watch whether any financial regulatory body (FATF, FinCEN, or an EU AML authority) cites or pilots a TVA-style scoring mechanism within the next 18 months. Adoption at even one national regulator would signal the framework is moving from theory toward enforceable policy.