딥러닝과 데이터마이닝에 있어서의 데이터베이스권의 범위와 제한에 관한 고찰

Abstract

As deep learning and data mining show astonishing achievements in certain areas, they are expected to open a bright future that we never anticipated before. However, they will probably conflict with the Copyright Act as copying of massive data is inevitable to implement these new technologies. There have been voices that the rights under the Copyright Act should be restricted to avoid chilling effects on these new technologies. Such arguments seemingly overlook that the policy of technological advancement should have a firm ground on the balance between the exclusive rights and the interest of the public in using the works. Without careful study and verification of the current balance, driving technological advancement would hardly lead to a well-balanced reading of the statutes. It would more likely bring more confusion by shifting the equilibrium as the need arises. Sui generis database rights, which are the main subject of this paper, do not require creativity as copyrights to works of authorship do. Instead, databases must have an arrangement or structure that enables access and search of individual data. This characteristic of databases must be the core in understanding the statutory provisions to derive a logical and robust interpretation that is faithful to the language and consistent with the idea-expression dichotomy prevailing throughout the Copyright Act. After an analysis based on the proposition above, this paper suggests that a “reproduction of a database” under Article 93(1) of the Copyright Act must include at least a part of the arrangement or structure in the original database, as copying a work of authorship requires copying the expression in the original work. In addition, a “systematic and repetitive reproduction of individual data” should enable access and search that is at least partially equivalent to those provided by the original database to “conflict with the ordinary use of the database” or to “unfairly infringe on the interest to the database” as stipulated in Article 93(2) of the Copyright Act. According to the interpretation suggested above, collecting data for deep learning and data mining in most cases would not be subject to the database rights or exempted under the law of fair use. That said, introducing new statutory provisions that incorporate the discussion in this paper is still relevant as they would reduce ambiguities. The Database Directive (96/9/EC) has been maintaining relatively narrow requirements for restriction of rights until amended by the Digital Single Market Directive (Directive (EU) 2019/790) recently. As a result, we do not yet see any amendments in European countries noteworthy for the legal situation in South Korea. Japanese Copyright Act has more advanced provisions restricting copyrights to allow new technologies such as deep learning and data mining, which apparently influenced the bill for the amendment of the Copyright Act in South Korea. However, the bill does not seem to address the issues properly to achieve the desired outcome. The new provisions should have clearer requirements that are formed around the ‘expression’ for works of authorship and the ‘arrangement or structure’ that enables ‘access and search’ for databases.