Artificial Intelligence Governance of Telemedicine Triage Chatbot Models Under Unmeasured Facility Ownership Conflicts in Provincial Licensing Boards: A Healthcare Political Capture Estimation

Provincial licensing boards that approve telemedicine triage chatbot models frequently contain members with undisclosed ownership stakes in facilities whose patient flows those models govern. The indicator has a mean of 12.674 and a standard deviation of 6.52, with country values ranging from 5.70 in Morocco to 27.10 in Malawi. The framework treats board composition as a latent variable that biases model approval thresholds, derives a capture bound from observable approval patterns, and specifies the conditions under which the bound is informative. Ownership-conflicted boards systematically relax triage thresholds for facilities they own, producing a measurable divergence between approved model behaviour and population-level need. The contribution is a formalisation of capture as a threshold-shifting mechanism rather than a binary corruption event, together with an estimation strategy that does not require access to board ownership records. The paper concludes that governance reform should target disclosure rules and threshold-setting authority rather than model auditing alone.