生成式人工智能开发的版权风险与化解路径——以ChatGPT为例

Copyright Risk and Solution of Generative Artificial Intelligence Development—Take ChatGPT as an Example

生成式人工智能拓展了文本生成的方式,提高了内容的产生效率并促进了专业知识的可及性,具有被制度鼓励的技术正当性。生成式人工智能的开发需要复制大量作品作为训练机器学习的材料,但此种复制目前并不受我国著作权立法的豁免。文本与数据挖掘过程中对作品的利用不以欣赏表达或获取信息为目的,不会对著作权人的市场产生替代,具有被作为权利例外的理论基础。域外版权立法也承认文本与数据挖掘中作品利用行为的合法性。未来著作权立法的修改,应当在合理使用条款中增加文本与数据挖掘利用作品的权利例外,对使用目的、使用行为和使用结果作出具体的规定。为应对立法权利例外规定的缺失,应充分利用著作权内置的私人自治、利用默示许可、探索与出版商合作和构建共享数据库等替代性措施来规避当前作品利用中存在的风险。

Generative artificial intelligence expands the way of text generation,improves the efficiency of content generation and promotes the accessibility of professional knowledge,and has the technical legitimacy encouraged by the system.The development of generative artificial intelligence requires the reproduction of a large number of works as materials for training machine learning,but such reproduction is not exempted by copyright legislation in China at present.The use of works in the process of text and data mining is not for the purpose of appreciating expression or obtaining information,and will not replace the market of copyright owners,which has the theoretical basis of being an exception to rights.The extraterritorial copyright legislation also recognizes the legality of the exploitation of works in text and data mining.The amendment of copyright legislation in the future should add exceptions to the right of text and data mining to the fair use clause,and make specific provisions on the purpose of use,use behavior and use results.To address the absence of exceptions to legislative rights,mining practitioners should explore alternative measures such as working with publishers,building shared databases,and using temporary copying to circumvent the risks of current exploitation of works.