生成式人工智能的刑事犯罪风险与刑法应对  

Criminal Risk of Generative Artificial Intelligence and ItsCriminal Law Countermeasure

作　　者：黄明儒[1] 刘方可 Huang Mingru;Liu Fangke(Law Shool,Xiangtan University,Xiangtan,Hunan 411105)

机构地区：[1]湘潭大学法学院,湖南湘潭411105

出　　处：《河南财经政法大学学报》2024年第3期69-79,共11页Journal of Henan University of Economics and Law

基　　金：2019年国家社会科学基金项目“风险社会背景下行政犯扩张及其适用限缩研究”(项目编号:19BFX061);2022年湖南省社会科学基金项目“湖南新型犯罪形势与治理模式研究”(项目编号:22YBQ046);2023年湖南省研究生科研创新重点项目“共同犯罪违法所得共同退缴责任研究”(项目编号:CX20230541)的阶段性成果。

摘　　要：以ChatGPT为代表的生成式人工智能存在犯罪风险。爬取数据的行为存在侵犯数据安全的数据犯罪风险,应当扩展数据犯罪的行为方式和行为对象,在“侵入获取型”“其他技术手段获取型”基础上增加“爬取获取型”数据犯罪行为,从而形成三维行为规制格局,并取消对数据领域和数据类型的限制。因“调教失能”“AI越狱”等引发人工智能责任事故犯罪风险时,生成式人工智能提供者应当承担过失犯罪责任,但应当摒弃旧过失论与新新过失论,采纳新过失论作为判断依据。面对生成式人工智能生成内容存在的侵犯著作权的犯罪风险,刑法应当采取谦抑的态度,通过实质性智力投入与算法可解释性来认定生成内容的作品性。The generative artificial intelligence represented by ChatGPT has a criminal risk.The behavior of data crawling poses a data crime risk of infringing on data security.It is necessary to add the criminal behavior of “data crawling” on the basis of “access by intrusion” and “access by other technicalmeans” so as to form a three-dimensional behavior regulation pattern.As the object of data crime,the restrictions of computer information system on data should also eliminated.When the risk of AI liability accident is caused by “training disability”,“AI prison break”,etc.,the generative AI providershould bear the responsibility for negligence crime,but should abandon the old theory of negligence and the new new theory of negligence,and adopt the new theory of negligence as the judgment basis.In the face of the copyright infringement risk of the generated content by generative artificialintelligence,the criminal law should take a modest attitude and identify the originality of the generated content through substantive intellectual input and algorithmic interpretability.

关 键 词：ChatGPT 生成式人工智能 数据犯罪 人工智能责任事故犯罪 知识产权犯罪 

分 类 号：D924.1[政治法律—刑法学]