挑战与应对:大型语言模型(ChatGPT)的多样态安全风险归因及协同治理研究

Challenge and Response:Research on Diversified Security Risks of Large LanguageModel(ChatGPT)and Collaborative Governance

ChatGPT带来大型语言模型的时代变革。ChatGPT的诞生是人工智能机器人取得的一个阶段性进步,其作为科技与日常生活融合背景下的突破性成就,对其引发的安全风险必须引起社会各界的足够重视。文本梳理了ChatGPT带来的安全风险,在人工智能治理基本框架的基础上,展开基于风险议题、治理主体和治理手段的三维分析,提出应对大型语言模型安全风险的治理原则和思路,并建议从法律规范、协同治理、部门监管、技术攻关和伦理约束五个维度,推进以ChatGPT为代表的人工智能安全风险治理。

ChatGPT brought about a change on large language model.The birth of ChatGPT is a phased progress of artificial intelligent robots.The security risks it brings about must be paid enough attention to.This paper analyzes the security risks brought by ChatGPT.Three types of risks caused by ChatGPT are summarized.Based on the theoretical analysis of AI risk and governance,a three-dimensional analysis framework is established to make a comparative analysis of ChatGPT risk governance.Finally,suggestions are put forward from five aspects:legal norms,collaborative governance,departmental supervision,technical research and ethical constraints.