生成式人工智能生成图片的著作权保护研究  

On Copyright Protection of Pictures Generated by Artificial Intelligence

作　　者：谭佐财 王梓隆 Tan Zuocai;Wang Zilong

机构地区：[1]华中科技大学法学院 [2]华中科技大学人权法律研究院

出　　处：《电子知识产权》2025年第2期15-26,共12页Electronics Intellectual Property

基　　金：华中科技大学人权法律研究院研究专项课题“数字时代财产权保护机制的构建研究”(项目批准号:2024RQJD001)的阶段性成果。

摘　　要：生成式人工智能生成图片已经成为不容忽视的新兴科技现象,其在一定程度上著作权法上的法律适用难题。生成式人工智能由于脱离“意志—理性”的人格分析框架而不具有实际的人格主体地位。同时基于法的实证规定以及契约学说的公意论原理,其也无法类比法人获得“拟制人格”。对生成图片的作品认定不宜采取固定标准,而应结合创作投入、指令输入程度、人工智能的精细化修改功能等因素综合认定。从权利类别角度审视,人工智能生成的图片因缺乏“审美”要素而无法被认定为著作权法中的“美术作品”,应纳入“其他类型”。依照媒介理论,人工智能对数据库的训练可以解释为使用者大脑思想之延伸,因而其生成的作品也应归属于使用者。软件使用者若采用“图片输入”的训练方式则应结合使用者的使用目的、生成用途等认定行为是否构成侵权。在责任分配方面,软件使用者承担过错责任,软件所有者承担无过错责任,而软件的设计者应承担相应的补充责任。The image generated by generative artificial intelligence has become a new technology phenomenon that can not be ignored,which causes the problem of legal application in copyright law.Due to its separation from the personality analysis framework of"volition-reason",generative artificial intelligence does not have the status of actual personality subject.Meanwhile,based on the positive provisions of law and the principle of the general will theory of contract theory,it cannot obtain"fictitious personality"by analogy with legal person.It is not appropriate to adopt fixed standards for the identification of works that generate pictures,but should be combined with factors such as creative input,instruction input degree,and refinement modification function of artificial intelligence.From the perspective of rights category,the pictures generated by artificial intelligence cannot be identified as"works of art"in the copyright law because of the lack of"aesthetic"elements,and should be included in"other types".According to media theory,the training of the database by artificial intelligence can be interpreted as an extension of the user’s brain,and the resulting works should also be attributed to the user.If the software user adopts the training method of"image input",he should determine whether the behavior constitutes infringement according to the user’s purpose of use and the purpose of generation.In terms of liability allocation,the software user bears the fault liability,the software owner bears the no-fault liability,and the software designer should bear the corresponding supplementary liability.

关 键 词：生成式人工智能 生成图片 媒介理论 美术作品 合理使用 

分 类 号：D923.41[政治法律—民商法学]