全球人工智能风险治理的趋势与特征——基于25份政策文件的文本量化分析

The Trend and Characteristics of Global AI Risk Governance--Text Quantitative Analysis Based on 25 Policies

人工智能已经成为新一轮科技革命和产业变革的重要驱动力量。选取2016年以来全球各国或组织出台的25份人工智能风险治理政策展开文本量化研究,构建“风险议题—治理主体—政策工具”三维分析框架,分析全球人工智能风险治理政策的主要内容、基本特征和发展趋势后发现,全球人工智能风险治理政策侧重于解决社会风险问题,对技术风险和法治风险的关注仍待强化;治理主体多样化,但参与度不均衡;以强制型政策工具为主,引导型、市场型政策工具为辅,自愿型政策工具运用偏少。因此,应优化政策工具整体布局,形成多元主体参与协调机制,健全人工智能安全等级和风险预估机制。

Artificial intelligence has become an important driving force for the new round of scientific and technological revolution and industrial transformation.After electing 25 AI risk governance policies issued by countries or organizations around the world since 2016,a text quantitative research was carried out.Through constructing a three-dimensional analysis framework of“risk issues-governance subject-policy tools”,and analyzing the main content,basic characteristics and development trend of global AI risk governance policies,it is found that the global AI risk governance policy focuses on solving social risks,and the attention to technical risks and legal risks still needs to be strengthened;The governance subjects are diversified,but the participation is uneven;Compulsory policy tools are mainly used,supplemented by guiding and market policy tools,and voluntary policy tools are less used.Therefore,we should optimize the overall layout of policy tools,form a coordination mechanism of multiple subject participation,and improve the AI security level and risk estimation mechanism.