人工智慧生成作品之著作權問題研究

人工智慧的浪潮在這幾年席捲全世界，在這幾年全世界也開始對於人工智慧有越來越多法律上的爭議，其中包含對著作權、專利權制度的衝擊、人工智慧的民事責任、個資保護的相關議題等等，而本文欲聚焦討論之問題即為人工智慧經訓練後而可進行創作之著作權問題。在世界各國，尤其是美國、歐洲、澳大利亞等國，自電腦問世以來，便對於電腦生成之作品有為數眾多之探討，然而技術進步快速，人工智慧生成之作品於技術上已與傳統電腦生成作品有所差異，因此，本文對於現今之人工智慧之技術亦會進行相關部分的介紹。本文乃是採取人工智慧進行開發時按照順序所發生的風險及問題做為思考順序：率先可能發生的問題為人工智慧進行訓練時所可能產生的侵權風險，然而目前並無任何國家承認人工智慧具有法律人格而得享有法律上權利義務，從而前述之侵權風險應由何人承擔？本於此問題，本文認為必須回頭探討最根本之問題，即人工智慧輸出之作品是否具有著作權保護？若無，則可能不生前述侵權之問題，若有，則須進一步討論權利歸屬之問題，並由權利人承擔該侵權風險。當然，本文之討論不僅僅限縮於侵權風險之問題，整體之重點仍然在人工智慧生成作品所產生之可著作性以及著作權歸屬問題，因此在架構之敘述上，仍然會按照一般討論法律概念之順序進行討論。在法律及政策分析之部分，本文將會採用文獻分析法，以文獻回顧之方式，系統性地整理美國、歐盟、英國、澳大利亞等國相關之論著與判決，並進行相關理論之解釋，以及比較法之分析與撰寫，試圖對人工智慧生成作品之可著作性以及著作權歸屬之問題進行梳理與分析。最後本文會對於上述比較法分析做出總結並試圖參考各國現行法下對於人工智慧生成作品的解決方案，以做為我國未來立法參考之用。

關鍵字

人工智慧；人工智慧生成作品；電腦生成作品；機器學習之侵權問題；著作權

並列摘要

In recent years, our lives are filled with artificial intelligence systems. Artificial intelligence systems are way better than the concept “computer” that people used to know; they are creative, independent, autonomous, rational, evolving. As a result, the world has begun to have more and more legal disputes over artificial intelligence. In all those legal issues, this paper will focus on the copyright issues related with artificial intelligence-generated works. Since it is way too easy to let a well-trained artificial intelligence generate creative works independently, the use of artificial intelligence, exercising varying degrees of autonomy, in the production of works, either for personal or manufacturing purposes, has become common. Despite this progress, there is a concern in modern society about artificial intelligence technology might getting out of control. Therefore, there is a call for social and legal tools for controlling artificial intelligence systems’ functions and outcomes. This paper addresses the question of the risk of infringement that may occur when training artificial intelligence, the copyrightability of artificial intelligence-generated works, and who should enjoy the benefits of copyright protection for, as well as be responsible for the infringement of rights and damages caused by artificial intelligence generated works. This paper presents several options, arguing against the imposition of these rights and responsibilities on the artificial intelligence themselves, developers, or on the user of the artificial intelligence systems. In the conclusion part, this paper will summarize the above-mentioned analysis, and also suggest some possible ways of reform warranting deeper exploration.