Artificial Intelligence Governance of Pandemic Contact-Tracing Models Under Unmeasured Subnational Data-Sharing Refusal Authority: A Healthcare Political Privacy Evaluation

Subnational authorities in African health systems frequently refuse to share data with the central state, yet the magnitude of this refusal is rarely measured and almost never modelled. The framework distinguishes three refusal regimes—latent, declared, and operational—and specifies how each distorts the training distribution, the privacy calculus, and the accountability chain of pandemic contact-tracing models. Using World Bank World Development Indicators for 24 African countries over 2000–2023 (N = 560 country-years), the paper estimates a cross-country model of GDP per capita on electricity access and urbanisation to characterise the infrastructural gradient along which refusal authority is exercised. The OLS estimates (R² = 0.591, F = 184.764) show that electricity access (coefficient 1291.554, SE 194.352) and urbanisation (coefficient 113.140, SE 7.775) are strongly associated with income, while the residual infrastructural term (coefficient −104.418, SE 11.313) captures the unmodelled variation that refusal authority occupies. Governance of contact-tracing AI must treat refusal as a legitimate jurisdictional act, not a data-quality defect, and proposes a refusal-aware audit standard that makes the unmeasured measurable.