Abstract:The adjustment of copyright has always been accompanied by the development of high technology. The development of artificial intelligence (AI), represented by ChatGPT and GPT-4, has brought great challenges to the copyright law system. Around issues such as the copyright of AI-generated content and the need to revise the criteria for originality, scholars have had a debate from different starting points and research frameworks, and formed completely different points of view. There is an urgent need to clarify the copyright rules, value orientation and coping strategies behind those technologies.As for the legitimacy of empowerment of AI-generated content, the paper believes that AI is not so much of a tool and it has further got rid of human arrangement. AI has directly intervened in the field of creation that was previously thought to belong exclusively to human beings, and can create algorithms independently and randomly to generate content that looks no different from ordinary works. In terms of originality, AI-generated content still falls within the framework of copyright law about protecting the expression of human thoughts and emotions, and it can meet the standard of independent completion and minimum creativity. As an important legal system for safeguarding the achievements of scientific and technological development, copyright law needs to respond to the iterative needs of AI, carry forward the ethical orientation of Tech for Good, and recognize the differences between algorithm-based creation and human creation. Copyright law should prioritize the protection of readers’ reliance interest in the appearance of AI-generated content, thus promoting the future development and transformation of AI.As for the evaluation standard of the consistency of subject and object, the paper holds that the identification of the works and the recognition of the subject belong to different dimensions. However, this does not conflict with the principle of private law that subject and object cannot be interchangeable. In the context of reader-centrism, the works as the object and the right holder as the subject can be appropriately separated. The works are not a natural extension of the author’s personality, and who the author is no longer affects the judgment on whether it constitutes the expression of ideas and originality. The theory of legal fiction provides a realistic example for the dichotomy of work creation and ownership, and the commercialization trend of works accelerates the process of separate evaluation of works and right holders.As for the arrangement of the right ownership of AI-generated content, it cannot impact the subject structure of the copyright and will not make AI the subject of civil law. In the case that AI-generated content constitutes works, we can still follow the idea of dividing the qualification of the creative subject and the qualification of the right holder. That excludes the possibility of AI as the copyright owner. Meanwhile, the real right holder can be defined according to the rules determined by current copyright practices. Compared with the creator and owner, the user shows a closer relationship with algorithm-based creation, which plays a decisive role in the content and form of AI-generated works. It is suggested that according to the actual contribution to creation, AI users should be regarded as the original right holders of the generated content. The basic conditions of copyright belonging to users are set from three levels: reverse exclusion elements, positive constitutive elements and formal description elements.As for the future of copyright in the era of intelligence augmentation, the paper points out that human beings must adhere to the principle of the unity of instrumental rationality and legal rationality while embracing the tide of the times with evolutionary ideas. We ought to accommodate human practices with more open concepts of the object and the author.
黄云平. 人工智能生成内容的可版权性问题辨析[J]. 浙江大学学报(人文社会科学版), 2024, 54(2): 75-90.
Huang Yunping. An Analysis of the Copyrightability of AI-Generated Content. JOURNAL OF ZHEJIANG UNIVERSITY, 2024, 54(2): 75-90.