Governments Have Moved Past Asking Whether AI Belongs in Healthcare. Now They Have to Figure Out How to Govern It.
AI policy is rapidly moving from principles to implementation. The difficult work now sits at the intersection of regulation, procurement, interoperability, workforce adoption, patient safety, equity, and accountability.
When representatives from 37 countries gathered to work on AI governance in health, the conversation had already moved past the question that dominated policy discussions three years earlier: whether AI belongs in clinical and public health systems at all. The active question now is how to govern something that is already being procured, deployed, and used in clinical decision-making across health systems with wildly varying regulatory capacity.
Principles are abundant. Implementation infrastructure is not.
Most countries now have some version of an AI-in-health governance principle document. Far fewer have the regulatory capacity to actually evaluate a clinical AI tool before it's deployed, the procurement expertise to negotiate vendor contracts that protect patient data and system interoperability, or the workforce training pipeline to help clinicians use these tools appropriately rather than either over-trusting or ignoring them.
The equity question is becoming a governance design question
Lower-resource health systems face a specific risk: adopting AI tools built and validated on data from wealthier populations, then discovering performance gaps only after deployment. Getting governance right means building in local validation requirements and interoperability standards before procurement, not auditing outcomes after the fact. That's a technical and institutional design challenge most health ministries are not yet staffed to handle on their own.