生成式人工智能犯罪中研发者刑事责任的认定——以ChatGPT为例

Determination of criminal liability of developers in generative artificial intelligence crimes——Taking ChatGPT as an example

生成式人工智能ChatGPT给社会带来福祉的同时,也蕴含着潜在的刑事风险。对涉及ChatGPT等生成式人工智能犯罪中研发者的不法与责任进行判断,有助于明确刑法介入人工智能技术风险的边界。涉及ChatGPT等生成式人工智能犯罪中研发者行为不法的确定,核心在于研发行为与法益侵害结果之间是否具有刑法上的因果关系。客观归责理论中是否创造并实现法不允许的危险,可以作为判断法益侵害结果是否属于研发者“作品”的标准。而ChatGPT等生成式人工智能犯罪中研发者责任问题的解决,则主要涉及研发者主观罪过的判定。当法益侵害结果出现时,研发者以引发社会秩序紊乱为初衷,将反社会语言程序注入ChatGPT中,应认定其具有直接故意。如若研发者以促进科技进步为初衷,未将反社会语言程序注入ChatGPT中,研发者主观罪过是否为过失取决于其是否履行了相应的注意义务。

Generative artificial intelligence ChatGPT not only brings benefits to society,but also contains potential criminal risks.Judging the illegality and responsibility of the developers of ChatGPT involved in generative artificial intelligence crimes can help clarify the boundaries of criminal law intervention in artificial intelligence technology risks.The determination of the illegality of the developer in generative artificial intelligence crimes involving ChatGPT and others lies in whether there is a criminal causal relationship between the research and development behavior and the result of legal interest infringement.The objective attribution theory can be used as a criterion to determine whether the result of legal interest infringement belongs to the“work”of the developer,whether it creates and realizes risks that are not allowed by law.The resolution of the responsibility issue of the developers of ChatGPT in generative artificial intelligence crimes mainly involves the determination of subjective guilt of developers.When the result of infringement of legal interests appears,the developer should be deemed to have direct intent when injecting anti social language programs into ChatGPT with the intention of causing social disorder.If a developer,with the original intention of promoting technological progress,fails to inject anti social language programs into ChatGPT,whether the developer’s subjective guilt is negligence depends on whether they have fulfilled the corresponding duty of care.