Purpose/Significance To explore the compliance obligations and risk mitigation strategies relevant to medical corpora in generative artificial intelligence (GenAI) in China, and to promote the compliance construction of corpora in this field. Method/Process Based on the data lifecycle of corpora in GenAI, the compliance obligations of service providers are sorted out by step. Targeted solutions are proposed to address three major risks:privacy leakage, training dataset bias, and intellectual property risks. Result/Conclusion It should be emphasized legality as the foundation, following key principles such as purpose limitation and data minimization, data rights protection, data security and risk prevention and control. Automated auxiliary means should be adopted to address various risks faced in the construction and management of medical corpora in GenAI.