大型语言模型的法律风险与治理路径

Legal Risks and Governance Paths of Large Language Models

大型语言模型促进了生成式人工智能的蓬勃发展。大型语言模型通常作为通用生成式人工智能的技术底层发挥作用,具备信息价值、情感价值、思维价值和劳动价值,也包含信息内容风险、数据与算法安全风险、侵权风险、诚信风险等系列法律风险。大型语言模型及相关应用的法律风险治理面临微观和宏观层面的系列挑战,需要平衡多方面的目标,应当基于包容审慎监管、分层治理、深度治理和敏捷治理的路径对大型语言模型的法律风险治理开展谨慎、灵活的探索性制度实践,进而逐步完善生成式人工智能的风险治理体系。

Large language models have facilitated the flourishing of generative artificial intelligence.Large language models usually play a role as the technical foundation of general'-purpose generative artificial intelligence which is enabled to offer information value,emotional value,thinking value and labor value.However,they entail a series of legal risks that influence information content,data and algorithm security,infringement,and integrity.The legal governance of the risks rendered by large language models and related applications faces challenges at micro and macro levels with the necessity to balance multi-faceted goals.Prudent and flexible exploratory institutional practices for the legal risk governance of large language models should be carried out based on the paths of inclusive prudential supervision,hierarchical governance,in-depth governance and agile governance,consequently the risk governance system of generative artificial intelligence can be gradually improved.