Microsoft's AI, The Next Rembrandt, learned from the paintings of 17th-century Dutch painter Rembrandt through AI learning and created AI creations with the same style. Some say this is plagiarism and reproduction, not original work.

Recently, image generation applications utilizing AI technologies such as deep learning and machine learning are being actively developed, and prototypes are being released to the public. As concerns and expectations coexist in the illustrator industry ecosystem, copyright law issues regarding commercial use are expected to surface.

■ Japan's Machine Learning Paradise Shows Off Bold Copyright Law Reform
Artificial intelligence technologies such as machine learning and deep learning are considered next-generation industries and are expected to become the center of cutting-edge industries. In particular, Japan has shown strong openness and reform movements in its copyright law, and has shown the most digital and AI-friendly deregulation among copyright laws in the world.

On the 14th, the 14th Korea-Japan Copyright Forum was held at the InterContinental Seoul COEX in Samseong-dong. At this forum, co-hosted by the Ministry of Culture, Sports and Tourism and the Agency for Cultural Affairs of Japan and organized by the Korea Copyright Commission, the topic of “Japanese Copyright Policy in the Era of the Fourth Industrial Revolution” was shared about copyright law revisions and issues that formed the basis for establishing a machine learning (ML) industry environment, as well as global trends.

Professor Tatsuhiro Ueno of Waseda University's Faculty of Law attended the forum that day as a presenter and gave a presentation on the 'Progress of the Exemption Regulations for Activating the AI and Big Data Industries', asserting that Japan has become a machine learning paradise through its preemptive revision of copyright laws.

Japan has the world's most advanced regulations for big data utilization and AI development due to the rights restriction regulations introduced in 2009 and revised and expanded in 2018. Article 30, Paragraph 4, Item 2 of Japan's Copyright Act, revised in 2018, states that "if the purpose is not to enjoy emotions yourself or have others enjoy them, you may use them in any way you deem necessary."

When applied to actual fields, the use of information analysis such as text data mining, big data analysis, and machine learning is interpreted as not being 'enjoyed' by oneself or a third party. However, a proviso provision stipulates that use is not permitted only when it unfairly infringes on the interests of the copyright holder.

Professor Ueno gave examples, saying, “This regulation enables various machine learning, and it enables the development of applications such as AI modeling that learns from many works to create new content, and learning from a large number of documents to produce results.”

According to Japanese regulations, it will be possible to explore new businesses in the future by utilizing AI, machine learning, and deep learning. Through text data mining, it is possible to analyze SNS texts such as Twitter to predict items that will be popular in the next week, and to report on trending issues. In the more distant future, it may be possible to develop AI new drug recipes by learning medical papers. Professor Ueno mentioned this vision and added that the problem of copyright use in machine learning should be resolved first.

■ Korean Machine Learning-related Copyright Law, Discussed for Revision for Several Years

In Korea, there have been several discussions on copyright law for the 4th industrial revolution, including AI, machine learning, and big data, but no visible revisions have been made to date.

Machine learning and deep learning require data sets for learning, but it was unclear whether using copyrighted data, such as pictures or texts by specific artists, for learning without the permission of the copyright holder would violate copyright law. As a result, commercial use of related services had no choice but to shrink due to the possibility of copyright infringement.

While the entire revision bill (proposed by Rep. Do Jong-hwan) proposed in 2021 failed to pass the legislative threshold, attention is being paid to whether some of the recently proposed revision bills will pass. Recently, Rep. Lee Yong-ho and 14 others proposed a partial revision bill to the Copyright Act, seeking to ease copyright regulations related to large-scale information analysis technology using artificial intelligence and big data technology. It is a basic regulation for protecting the rights of copyright holders and activating the AI industry as technology advances, while clarifying the boundaries of copyright infringement in machine learning.

According to the proposed amendment, data that has been approved by the copyright holder can be freely used for machine learning, etc., but necessary measures prescribed by Presidential Decree, such as measures to prevent duplication, must be taken. It is stated that the data can be used for non-commercial purposes such as education, surveys, and research, and for the purpose of creating works.

Unlike Korea's information analysis copyright policy, which is still in its infancy, Japan boasts of being a machine learning paradise, and is showing off its 'spicy' openness.

In Japan, there are no restrictions on profit or non-profit in information analysis, and even 'opt-out' is not possible, so even data that the copyright holder has refused to collect can be used. It is said that even content obtained from illegal sites can be used because there are no legal access requirements for data access.

It seems that the reason such strong regulations are possible is because the interpretation of copyrighted data to create new AI creations is not considered an infringement on the enjoyment of the original work, nor is it an infringement on profits. Professor Ueno advocated for the opening of copyright for machine learning, saying, “Information analysis is not the use of ‘information’ itself.”

■ “The Fall of the Creator” vs. “The Change in Creative Tools” At the forum that day, we were able to get a glimpse into the perspectives of experts and officials on the future that AI creative applications will bring. There were mixed views, such as concerns that this would lead to a decline in the motivation and downfall of creators, and views that it was simply a paradigm shift in creative tools, like the period when paintings were replaced by cameras.

Tsukasa Aso, an associate professor at Kyushu University Graduate School of Art and Engineering, said in a presentation on the protection of rights for AI-created works that “AI-created works are not protected because they have difficulty meeting the requirements of ideas and emotions in copyright.”

Japan is already leading the way in discussions on the rights of creators and rights holders regarding AI creations, and is preparing for the next step by discussing issues such as ▲rights holders, ▲false indication of authors, and ▲relevance for determining copyright infringement relationships between AI creations and the original works that trained them, based on the theory of protection and legislation for AI creations.

Regarding the rights holder, Professor Aso said, “I do not think that copyright can be recognized with just a one-line instruction to draw a picture or a user clicking a button to create a picture,” and that recognition of creation through specific instructions and additional work is necessary, and also opened up the possibility that the AI program creator could become the rights holder.

In the Q&A, one panelist advocated for recognizing the rights of creators involved in AI creations, saying, “Just as we moved from the era of paintings to the era of photography, AI is just a tool, and the people who use the tool should also be given rights.”

On the other hand, another panelist said, “If creators are positioned as lower-ranking members of the industry in the distribution process of AI products, there are concerns that this will lead to the devastation of the human spirit and the proliferation of mechanical creations that value the results more than the creative process.”

In this process of AI technology development, it seems that the key will be to control the speed between protecting copyright holders and forming an industrial base. Looking at Japan's progress, it is expected that issues such as ▲establishing a copyright incentive system for machine learning data providers and ▲regulating the rights to copy and protect AI products will emerge as controversial issues in the future.