Abstract: Generative AI models represented by ChatGPT, Sora, etc. have strong content understanding and creation capabilities, which will empower intelligence production and utilization in many application scenarios and contribute to the change of intelligence paradigm. Due to algorithmic and data security risks caused by the technical defects of generative AI, subject crisis and trust crisis caused by the decline of AI objectivity, and imbalance of personal information protection and responsibility determination caused by the loopholes in institutional norms, key national intelligence fields such as scientific and technological intelligence, military intelligence, financial intelligence, and public security intelligence are facing serious security challenges. It is of practical significance to build a layered governance system of “technology, ethics and system” to innovate the governance paradigm of national security intelligence work. At the technical level, strengthen the research of independent technology and enhance the risk prevention capability to optimize the technical model and security of intelligence system; at the ethical level, improve the subjective awareness of intelligence personnel and strengthen the human-machine collaboration mechanism to reduce the risk of intelligence research and decision-making; at the institutional level, improve the construction of intelligent intelligence system, clarify the obligations and responsibilities of relevant subjects, and enhance the capability of intelligence supervision and regulation.