【摘要】 生成式人工智能(Generative AI)技术在提升档案智能化管理效率的同时，也带来了信息失真、算法偏见、数据安全、法律伦理等多维风险。为应对AI伴生风险，基于ISO 31000、AI RMF、等级保护2.0等标准及档案法规构建风险分级体系，以风险源、影响对象、损害程度和发生概率为核心评估维度，设计分级标准与评价矩阵以及指标筛选与权重赋值方法。基于风险等级划分，进一步提出分级应对策略以及融合PDCA动态循环治理与长期战略协同的风险治理机制。该体系为档案机构科学识别、评估和管理生成式AI应用风险，保障档案工作的安全、合规与可持续发展提供了理论框架与实践路径。更多还原

【Abstract】 While generative Artificial Intelligence(AI) enhances the efficiency of intelligent archival management, it also introduces risks such as information distortion, algorithmic bias, data security, and legal-ethical issues. To address these AI-associated risks, this study constructs a risk classification system based on standards such as ISO 31000, AI RMF, and Cybersecurity Classified Protection 2.0, as well as archival regulations. The system takes risk sources, affected objects, damage severity, and occurrence probability as core dimensions, and designs classification criteria, evaluation matrices, and methods for indicator screening and weight assignment. Based on the differentiated risk levels, graded response strategies are proposed, along with a risk governance mechanism that integrates the PDCA cycle with long-term strategic collaboration. This system provides a theoretical framework and practical pathway for archival institutions to identify, assess, and manage risks in generative AI applications, ensuring the safe, compliant, and sustainable development of archival work.更多还原