Scaling, Lock-In, and Proxy Compliance: A Political Economy of Responsible AI

This paper models AI accountability at scale as an institutional problem involving vendors, deployers, and regulators. Vendors choose auditability and substantive mitigation; deployers monitor systems after adoption while facing switching costs; and enforcement relies on verifiable evidence. The model predicts a “proxy-compliance” equilibrium in which vendors satisfy an observable procurement floor but underinvest in mitigation relative to the social first best. It analyzes how independent audit rights, portability, incident reporting, and outcome-linked liability alter enforcement exposure and incentives. The framework offers testable implications for the gap between formal compliance and post-deployment outcomes.