生成式人工智能嵌入数字政府的技术路径、潜在风险与制度规制  

Generative AI Embedded in Digital Government:Technical Pathways,Potential Risks,and Regulatory Framework

作　　者：程圆圆 CHENG Yuanyuan(Guanghua Law School,Zhejiang University,Hangzhou 310000,Zhejiang,China)

机构地区：[1]浙江大学光华法学院,浙江杭州310000

出　　处：《昆明理工大学学报（社会科学版）》2024年第6期19-28,共10页Journal of Kunming University of Science and Technology(Social Sciences)

基　　金：国家社会科学基金一般项目“公共数据授权运营的法律制度构建研究”(24BFX032)。

摘　　要：技术革新是政府实现治理能力现代化目标的重要推动力量。生成式人工智能技术具有通用性、生成性与交互性,能够胜任各种类型的任务,有利于促进行政组织的扁平化、降低行政成本、提升行政效率、推动科学决策与科学立法。具体而言,生成式人工智能可以赋能公众意见收集、政府信息公开制度,辅助或者独立作出具体行政行为,参与政策制定与行政立法。但生成式人工智能应用于数字政府领域可能导致如下三重风险:虚假信息生成与算法歧视风险;数据安全及个人信息保护风险;社会资本侵蚀公共权力风险等。为引导生成式人工智能“赋能”而非“负能”数字政府建设,降低技术应用风险,行政部门应当在部署前开展事前评估,进行成本收益分析,并且引入沙箱机制;在事中遵循技术性正当程序原则,开展全流程的数据安全保护;健全权利救济及追责机制。Technological innovation is an important driving force for enhancing governmental capacity in modern governance.Generative AI technology is versatile,generative,interactive and able to manage diverse tasks.It can help streamline administrative structures,cut costs,boost efficiency,and supports data-driven decision-making and legislative processes.Generative AI strengthens public opinion collection,enhances government information disclosure,assists in specific administrative tasks,and plays a role in policy-making and legislation.However,the use of generative AI in digital governance also poses three main risks of misinformation and algorithmic bias,threats to data security and privacy,and potential erosion of public authority by social capital.To ensure that generative AI empowers rather than undermines digital governance and reduce risks in technological application,agencies should take preventive measures.Administrative departments should conduct prior assessments and cost-benefit analyses and introduce sandbox mechanisms before deployment;follow the principle of technical due process during the event and carry out data security protection throughout the process;and improve the mechanism for accountability and rights remedies.

关 键 词：生成式人工智能 技术赋能 数据安全 信息风险 制度规制 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] D63[自动化与计算机技术—控制科学与工程]