[Purpose/significance] The widespread use of generative artificial intelligence has given rise to multiple da⁃ ta security risks. Traditional reactive and centralized governance models seem inadequate to address these new chal⁃ lenges. In contrast, agile governance, with its flexible approach, demonstrates a unique superiority. [Method/process] Firstly, a meticulous examination of the multifaceted risks in data security posed by generative artificial intelligence is conducted. Subsequently, a comparative analysis of the efficacy of three governance models in risk management is car⁃ ried out. It is suggested that China should promptly transition its governance model. Guided by this, a specific gover⁃ nance framework is constructed. [Result/conclusion] The agile governance approach, characterized by its adaptability, flexibility, and inclusiveness, offers a highly effective strategy for addressing data security risks associated with emerg⁃ ing technologies like generative artificial intelligence. It allows for the continuous optimization of the governance sys⁃ tem for generative artificial intelligence data security. Within the agile governance framework, it is essential to embrace the adaptive governance concept of "equal emphasis on prevention and response", construct a resilient governance mechanism that promotes "diverse participation and collaborative interactions"，and employ inclusive governance ap⁃ proaches that combine technology with legal measures. This way, a comprehensive framework for ensuring the security of generative artificial intelligence data is constructed.