Artificial Intelligence Governance of Cross-Border Clinical Model Sharing Under Unmeasured Data-Protection Regime Heterogeneity: A Healthcare Political Feasibility Evaluation

Cross-border sharing of clinical artificial intelligence models is governed through data-protection regimes whose heterogeneity is not measured by the institutions that must authorise it. the present analysis evaluates the political feasibility of such sharing by treating regime heterogeneity as an unmeasured structural condition rather than a residual risk. It develops a feasibility framework that separates three governance objects — model parameters, model behaviour and model provenance — and argues that each object encounters a different configuration of legal constraint, institutional incentive and accountability obligation. The framework is applied to the African regional context, where cross-border health data flows are governed by a patchwork of instruments whose substantive provisions diverge on lawful basis, secondary use and transfer safeguards. The analysis identifies a feasibility inversion: arrangements that appear most protective at the level of data transfer may be least protective at the level of clinical accountability, because they push model development toward jurisdictions with weaker oversight. Three governance propositions are derived concerning mutual recognition, behavioural attestation and provenance disclosure.