论生成式人工智能语料库数据收集风险与法治进路  

On the Risks of Data Collection in Generative Artificial Intelligence Corpora and the Path of Legal Governance

作　　者：许皓 刘婧 XU Hao;LIU Jing(School of Political Science and Law,Huanggang Normal University,Huanggang,438000,China)

机构地区：[1]黄冈师范学院政法学院,湖北黄冈438000

出　　处：《福建江夏学院学报》2024年第5期61-71,共11页Journal of Fujian Jiangxia University

基　　金：黄冈市市域社会治理研究中心开放基金项目“我国区域法治研究的演化轨迹和发展趋势”(202417704)。

摘　　要：生成式人工智能不管是通过直接提取或间接推理用户的提示文本,还是以恶意爬虫等不正当方式达到填充语料库目的,都可能外泄攫取到的个人、企业用户中的涉密信息。生成式人工智能语料数据库因其内部因素的不确定性和外部因素产生的诸多偏见,呈现出语料数据库“主动收集”和“被动收集”两种风险范式。基于“科技发展与监管策略”双元思路,实现生成式人工智能的语料数据库收集风险的有效治理应当遵循人本化的科技伦理,利用先进科技工具,从国家、行业和用户三个维度协同共治加以防范管控。Whether generative artificial intelligence achieves the purpose of filling the corpus by directly extracting or indirectly inferring the user's prompt text,or by improper means such as malicious crawlers,it may leak confidential information of personal and corporate users.Due to the uncertainty of internal factors and the many biases caused by external factors,the generative artificial intelligence corpus database presents two risk paradigms of"active collection"and"passive collection"of the corpus database.Based on the dual idea of"scientific and technological development and regulatory strategy",the effective governance of the collection risks of the corpus database of generative artificial intelligence should follow the humanistic science and technology ethics,use advanced scientific and technological tools,and coordinate and co-govern from the three dimensions of country,industry and user to prevent and control.

关 键 词：生成式人工智能 语料库数据 数据收集 数据治理 

分 类 号：D923.41[政治法律—民商法学]