In recent years, artificial intelligence (AI) has been widely used in the medical field. Research shows that medical AI can not only improve diagnosis and treatment efficiency, but also enable personalized health management and optimize the allocation of medical resources, demonstrating enormous potential. However, while “technology empowers”, medical AI is also accompanied by a series of ethical paradoxes, mainly manifested as the conflict between technological innovation and the protection of humanistic value. Based on the AI ethical framework proposed by Italian philosophy and ethicist Floridi, this paper systematically analyzes paradoxical manifestations of medical AI and the possible causes in 5 core ethical principles: respecting autonomy, not harming, doing good, fairness, and interpretability. It also explains the causes of these paradoxes from technical, social, and philosophical levels, and puts forward the following coping strategies: first, to improve explainability, algorithm audit technology, differentiated data collection technology, privacy protection technology, and human supervision and control technology, strengthening technical governance; second, to improve ethical regulations and establish a dynamic ethical governance mechanism that adapts to the pace of technological development, a hierarchical management model, and a responsible innovation mechanism, further improving the legal and regulatory frameworks in terms of data security, algorithm ethical responsibility recognition, etc; and third, to establish a global collaborative system, build a global governance framework for medical AI, establish unified development and application standards, and promote cross-border policy coordination and technical standard docking.