生成式AI刑事治理的应对思路与路径选择：基于ChatGPT模型的新挑-《集美大学学报》（哲社版）

文章信息/Info

Title:: Response Ideas and Path Selection of Generative AI Criminal Governance:New Challenges Based on ChatGPT Model

Author(s):: WU Peize; (School of Law and Politics,Yunnan University of Finance and Economics,Kunming 650221,China)

摘要:: 以ChatGPT为代表的生成式AI给人工智能领域的刑事风险应对带来了崭新挑战，对此，应采取从技术逻辑到规范评价的认定思路。ChatGPT的突破在于数据、算法、算力的共同进步。在刑法视域中，ChatGPT虽无法形成“自主意识”，从而拥有刑事主体身份，但呈现出特殊规制对象属性、犯罪工具属性与保护对象属性不断增强的基本态势。通过对主体范围、刑事风险种类、规制效果3个维度的检验可以发现，人工智能的刑事立法进路并不可取。立足于现有刑事规范并通过适度扩张其适用潜能与解释能力足够消解涉ChatGPT犯罪的归责困局。涉ChatGPT犯罪的刑事治理路径应围绕技术支配与应用场景的多元化予以构建。技术支配划定了归责的基本主体与可归责程度；应用场景的多元化则对归责主体的范围予以重塑，并发展出基于风险等级的风险控制模式与归责模式。

Abstract:: The generative AI represented by ChatGPT brings a new challenge to the criminal risk management in the field of artificial intelligence.In this regard,we should adopt the identification thinking from technical logic to normative evaluation.ChatGPT’s breakthrough was the convergence of data,algorithms and computing power.From the perspective of criminal law,although ChatGPT cannot form “autonomous consciousness” and thus possess the identity of criminal subject,it enhances the attributes of special regulation object,crime tool and protection object.Through the examination of the three dimensions of “subject scope”,“criminal risk type” and “regulatory effect”,it can be found that the criminal legislation approach of artificial intelligence is not advisable.Based on the existing criminal norms and by moderately expanding its applicable potential and explanatory ability,it is enough to resolve the hot phenomenon of ChatGPT related crimes.The criminal governance path of ChatGPT crimes should be constructed around “technology domination” and “diversification of application scenarios”.The former defines the basic subject and degree of imputation,and the latter reshapes the scope of attributive subject and develops risk control mode and attributive mode based on risk level.