人工智能服务提供者的版权侵权责任研究

Research on Copyright Infringement Liability of Artificial Intelligence Service Providers

对于以语言模型为代表的生成式人工智能系统而言,在认定其服务提供者版权侵权责任时,首先,应综合考虑人工智能的技术特征、行业监管等因素,将人工智能服务提供者作为新型网络服务提供者进行规制。其次,应基于全周期思路,从事前、事中、事后的内容生产全过程审视人工智能服务提供者的注意义务。从训练数据的透明度义务来看,人工智能服务提供者必须确保所使用的数据来源合法、透明,避免因数据侵权而引发版权纠纷。同时,生成过程的内容审核义务意味着服务提供者需要对生成的内容进行必要的审核,确保其不侵犯他人的版权。若最终生成侵权内容,服务提供者还需履行相应的必要措施,以规避侵权内容的扩大传播和再次生成。

For the generative artificial intelligence system represented by language models,when determining the copyright infringement liability of its service providers,firstly,the technical characteristics of artificial intelligence,industry regulation and other factors should be synthesized,and the artificial intelligence service providers should be regulated as a new type of network service providers.Secondly,the duty of care of AI service providers should be scrutinized based on the full-cycle thinking in the whole process of content production before,during and after the event.From the perspective of the obligation of transparency of training data,AI service providers must ensure that the source of the data used is legal and transparent,so as to avoid copyright disputes caused by data infringement.At the same time,the content audit obligation of the generation process implies that the service provider needs to carry out the necessary audit of the generated content to ensure that it does not infringe the copyright of others.If infringing content is eventually generated,the service provider also needs to fulfill the necessary measures to avoid the expansion of infringing content and its re-generation.