Abstract
This policy brief substantiates the legitimization of artificial intelligence (AI) in the healthcare system of the Republic of Kazakhstan (RK) based on the ST RK 4010-2025 standard. Following a comparative analysis of three approaches, the progressive-adaptive scenario is recommended. It is built upon the "Human-in-the-loop" principle and the continuous control of algorithm trustworthiness. This approach introduces advanced AI under the condition of subsidiary liability for software developers while preserving the final clinical decision exclusively for the physician. What is the problem? The integration of AI in Kazakhstan's clinics faces resistance from physicians due to legal risks and the "black box" effect—the opacity of algorithmic logic, which violates World Health Organization requirements and undermines clinical reasoning. The situation is aggravated by an educational vacuum: university curricula and professional development courses do not cover AI bioethics or the verification of cognitive computing. Without implementing standards for AI explainability, investments in digital medicine will remain purely formal. Policy options Option 1: Conservative Scenario: Maximum restriction of AI to simple static systems. Continuous learning is prohibited, and 100% of the legal liability for any system failures rests solely on the physician. This eliminates risks but leads to technological stagnation. Option 2: Progressive-Adaptive Scenario (Recommended): Creation of a flexible regulatory environment for the safe use of continuous learning AI. The prerequisite is the launch of a national system for monitoring trust in AI and the regular revalidation of models based on the Unified Medical Data Repository (UMDR). Option 3: Liberal-Technocratic Scenario: Rapid import of foreign autonomous AI to bridge the healthcare workforce deficit in the regions. AI systems are delegated the authority to automatically issue screening conclusions without a physician's signature. Vision on the implementation of policy options The progressive-adaptive scenario is optimal for the healthcare system of the RK. It distinguishes between process automation and system autonomy, legally securing the physician's priority in decision-making. The critical step is the establishment of a national system for monitoring trust in AI on the basis of the Salidat Kairbekova National Scientific Center for Health Development. This will link data collection within the UMDR and transform medical analytics into a safe foundation for clinical decision support systems. link data collection within the UMDR and transform medical analytics into a safe foundation for clinical decision support systems. Keywords: artificial intelligence, medical analytics, medical statistics, clinical decision support systems, data collection, data management, medical information systems.