生成式人工智能服务提供者中立参与行为的处罚范围  

Defining the Punitive Scope of Neutral Participation by Generative AI Service Providers

作　　者：黎森予 LI Senyu

机构地区：[1]清华大学法学院 [2]东京大学大学院

出　　处：《苏州大学学报（法学版）》2024年第4期40-52,共13页Journal of Soochow University：Law Edition

基　　金：国家社会科学基金项目“数据法益的刑法保护研究”(项目编号:20BFX074)的阶段性成果。

摘　　要：生成式人工智能服务提供者可能通过人工智能程序唆使、援助使用者实施犯罪,有必要探讨其处罚范围。一般而言,提供生成式人工智能服务的行为至多对使用者的犯罪承担共犯责任,注意义务违反这一过失犯的要素并不适用于共犯成立与否的判断;将故意作为界定处罚范围的标准,容易不当扩张处罚范围;中立共犯理论提出的限制处罚范围的各种标准,均存在疑问。应当基于“冲突性正当利益的衡量”这一超法规的违法阻却事由,判断提供生成式人工智能服务行为的可罚性:第一,行为必须促进、维护了正当利益;第二,行为促进的正当利益与其损害的法益存在冲突;第三,行为促进的正当利益不小于其损害的法益。即便提供生成式人工智能服务的行为与不当使用者造成的正犯结果有因果性,只要符合上述条件,就不具有违法性。Generative AI service providers may incite or assist users in committing crimes through their AI programs,which raises questions about the extent to which they should be held accountable.Generally,providing such services may at most lead to accomplice liability for the user's crimes,and the breach of duty of care does not apply to determining accomplice liability.Relying on intent to set the boundaries of punishment could lead to an unjustified expansion of liability,and the criteria suggested by neutral accomplice theories are often problematic.Instead,the criminal liability of providing generative AI services should be evaluated through the lens of“balancing conflicting legitimate interests,”an extra-legal ground for excluding unlawfulness:First,the conduct must promote or maintain legitimate interests;second,the legitimate interests promoted by the conduct must conflict with the legal interests it harms;and third,the legitimate interests promoted by the conduct must be no less than the legal interests it harms.Even if there is causality between the AI services and the principal offender's unlawful conduct,the provider's conduct should not be deemed unlawful as long as it meets the above conditions.

关 键 词：生成式人工智能 中立参与行为 中立共犯 利益衡量 违法阻却事由 

分 类 号：D924[政治法律—刑法学]