Artificial Intelligence Governance of Clinical Algorithm Registries Under Unmeasured Post-Market Performance Drift Reporting: A Healthcare Political Accountability Evaluation

Clinical algorithm registries are increasingly proposed as the institutional mechanism through which health systems will govern artificial intelligence at the point of care, yet the political accountability they promise depends on a reporting infrastructure that does not currently exist. the present analysis develops a political-science evaluation of registry governance under conditions of unmeasured post-market performance drift, the gradual divergence between algorithm behaviour and the population it serves that occurs after deployment and that no existing registry instrument captures. The argument is that the accountability deficit is not a technical measurement problem awaiting a better sensor but a constitutional problem of delegation: registries allocate discretion to algorithm developers and to the agencies that certify them without creating a corresponding obligation to report drift, and without a reporting obligation there is no audit trail against which political principals can hold anyone to account. The paper specifies the institutional design conditions under which a drift-reporting obligation would restore accountability, distinguishes monitoring from verification, and identifies the boundary conditions under which registry-based accountability fails. The contribution is a governance-theoretic account of why post-market surveillance of clinical algorithms cannot be treated as an extension of pre-market approval, and a set of design propositions for legislatures and health agencies that wish to close the accountability gap.