인공지능과 저작권 침해의 문제

Abstract

Creative works expressing human thoughts and emotions are recognized with exclusive rights known as copyright. With the recent advancements in artificial intelligence (AI) technology, the realm of thought and creativity, once considered exclusive to human intellect, has become possible for AI as well. This development requires substantial amounts of data, and since copyrighted material may be used during machine learning processes, discussions about access to and use of data, and the exceptions or immunities concerning Text and Data Mining (TDM), have emerged.First, the concept of data is examined and then expanded to discussions on data utilization. Subsequently, for the effective implementation of TDM exceptions—a topic of legislative interest in recent years—situations in major countries were compared. In the European Union, the Digital Single Market (DSM) directive, along with Directives 96/9/EC and 2001/29/EC, established exceptions for TDM. Exceptions or limitations are permitted for scientific research purposes, and guidelines for reproduction and extraction are stipulated for copyrighted works that member states can lawfully access for conducting scientific research. Article 3 provides immunity for subjects engaged in TDM for academic research and prohibits reservations on the application of this provision. However, Article 4 is interpreted as allowing all subjects and purposes without restriction, leaving room for the exclusion of its application by contract or opt-out.Japan, considered a leader in discussions on AI and copyright, has long debated the protection of AI-generated works. To promote investment in the AI industry and provide incentives for AI creation, Japan has categorized the forms in which AI is used in creation. It is suggested that if a person uses an AI system as a tool for creation, intellectual property rights should be granted to the person. However, if AI autonomously generates a work, it should not be subject to intellectual property protection and instead be considered public domain.In the United States, under Section 107 of the Copyright Act, activities may be exempt from copyright infringement if they qualify as fair use. Therefore, even if the data collected for big data purposes undergoes TDM and is interpreted in a different form, or if discussions arise regarding copyright infringement during the TDM process, it appears that such issues are judged based on the doctrine of fair use. When fair use applies, the use can proceed without the copyright holder's permission. Determining whether the use of copyrighted works for TDM constitutes fair use under Section 107 of the U.S. Copyright Act involves considering the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the portion used, and the effect on the potential market.Germany amended its copyright law in 2017 and introduced specific provisions for TDM. Article 60d of the Copyright Act, titled "Text and Data Mining for the Purpose of Scientific Research," specifies that research institutions may reproduce works for scientific research purposes to perform TDM. Japan's approach appears relatively liberal regarding TDM, as Japanese TDM is broader in scope than that of Europe. Japan distinguishes between the use of copyrighted works for enjoyment and non-enjoyment purposes, potentially allowing a broader exception for TDM. The TDM exception in Japan's legal system, rooted in the civil law tradition, is also evaluated as a potential alternative for expanding exceptions in the European Union.It is widely known that Article 35-5 of Korea's Copyright Act adopted provisions similar to Section 107 of the U.S. Copyright Act. However, given the limited accumulation of precedents, the criteria established in U.S. case law are still utilized as a standard for judgment. The provisions of Korea's Copyright Act, initially almost identical to Section 107 of the U.S. Copyright Act, have undergone significant changes, such as the removal of the phrase “for profit or non-profit.” There are calls for caution in applying the doctrine of fair use within our copyright law. Considering the development of generative AI, there is a need to gather expert insight to develop copyright law in a way that ensures predictability in applying the doctrine of fair use, protects copyright, and promotes the healthy use of big data through TDM exceptions. However, to enhance the effectiveness of TDM exceptions, it is necessary to clarify the scope and limitations of immunity provisions.