The large-scale application of artificial intelligence (AI) in the healthcare field has effectively enhanced diagnostic and therapeutic efficacy, optimized resource allocation, and provided robust support for precision medicine and public health prevention and control. Meanwhile, it has also given rise to a series of novel ethical and safety risks, including algorithmic bias, AI hallucinations, data privacy leakage, decision-making black boxes, and ambiguous responsibility boundaries. The core of medical ethics lies in the dignity of life, whereas AI ethics is targeted at risk controllability. The two are deeply intertwined in medical scenarios, presenting severe challenges to the traditional ethical governance system. This study systematically sorts out the policy evolution and current status of industrial practices related to healthcare AI ethics at home and abroad, analyzes the existing ethical challenges from seven dimensions, including technical robustness, privacy data governance, and fairness, and decomposes the AI-native risks that are difficult to be addressed by the traditional safety system. On this basis, a collaborative construction framework of human-machine alignment is established, which takes "safety, benevolence, trustworthiness, and controllability" as its objectives, covers the entire life cycle of the model, and integrates three layers of paths: technology, application, and norms. Furthermore, prospects are proposed in terms of technological R&D, system construction, talent training, and industrial development, thereby providing theoretical support and practical reference for the compliant application and standardized implementation of healthcare AI.
Ju et al. (Thu,) studied this question.