Developing Artificial Intelligence (AI) at the beginning was difficult due to absence of managing large amounts of data law. Then, AI started to produce unprecedented results of data acquisition via machine learning through the Internet. So, some people argue that we should grant copyrights of AI’s creations. However, there are various problems when granting copyrights of AI’s creations. Many artists can lose their jobs and it is impossible to grant copyrights of AI’s creations legally at the current situation. Moreover, many countries do not acknowledge copyrights of AI’s works. Therefore, the Korean government should not grant a copyright of AI’s creations.
First of all, numerous artists’ livelihoods can be at risk due to AI’s indiscreet ability to learn and generate work. Many artists expressed their opinions on this on ArtStation[1], the picture generating platform. The artists resist about uploading pictures made by AI to ArtStation. Nevertheless, AI’s images constantly appear on that platform which is deeply relevant to the livelihoods of many applicants and artists. Hence, the users who are developing their portfolio and sharing information about looking for jobs, expressed animosity to ArtStation that their positions are being threatened. Then, last December, they began to use a “No To AI” logo asking ArtStation to separate AI’s works from human made ones, or to ban AI images together. Especially AI with open-source program can bring additional economic problem as it gathers and learns from artwork without indicating the source or getting permission from the original authors. Also, according the Chosun Ilbo last October, “NovelAI” is producing qualified images for only about 15 won implying outstanding productivity. Moreover, AI can damage artists as it easily makes specific results with relatively greater speed than people can. For example, Greg Rutkowski, a Polish digital artist draws dreamy, classic landscapes. He uploaded his works with high-definition on the Internet to sites like ArtStation, and “Stable Diffusion[2]” collected his works without his awareness or approval. At first, he thought it as a good opportunity for people to become exposed to his art, but the Internet soon began to present works similar to his when searching for his name. So, according to MIT Technology Review in last September, he showed concerns that he can no longer find his own original works. He can lose his job since the company would use AI that can much more quickly produce images in his style. Thus, copyrights of AI’s creatures should not be granted.
Secondly, copyrights of AI’s creations cannot be admitted legally. According to the Article 2, Chapter 1 General Provisions of the Copyright Act of Korea, it says that “The term “work” means a creative production that expresses human thoughts and emotions; The term “author” means a person who creates a work;” In other words, “works” are the results of humans. Also, according to the Korean dictionary, “creative production” means “The overall results made by humans’ mental effort…” So, creative production of artworks by humans as subjects of the Copyright Act requires various effort including time, and technologies. However, AI generate results via machine learning, not with its own creativity. According to ekoreanews last December, a 3D artist said “[…] it is not right to treat the pictures by AI and the results of humans as the same.” Furthermore, in December 2020, the Korean government discussed over promoting AI to a legal entity, but failed to finish the discussion, leading to Evom, an AI composer, no longer receiving royalties since July 2022. In addition, in last November, there was a class action against “GitHub Copilot[3]” for generating the source code of Microsoft, which required compensation for damages from illegal copying of software in the United States. This means that results by mixing learned works, not originality became problematic in this case. Thus, it is impossible to grant copyright to AI as the creations via this process are not correspondent with the concept of work of the Copyright Act.
Thirdly, most countries do not grant copyright to AI creations. In February 2022, doctor Stephan Thaler developed the “Creativity Machine” and it made the work “A Recent Entrance to Paradise.” He then tried to register for copyright, but the U.S. Copyright Office denied his request saying “That work failed to fulfil the requirement of human author.” Also, in December 2021, RAGHAV, an AI painting application, was acknowledged as a joint author, not as a solo author of art works in Canada. This is because that the Canadian Copyright Law implies human authors by counting the copyright term after the authors’ deaths. In other words, it is unusual to grant a copyright to only AI’s creations. However, another problem can appear even when many countries grant the solo copyrights for AI: ownership of royalty. For example, when AI generated a sound source, does the ownership of the sound source belong to the production firm that created the AI, or the person who intended to make the sound source? Such vagueness explicitly demonstrates the difficulty in recognizing the copyright for AI’s creations, even if the copyright is globally granted.
Rutkowski said “These changes are wonderful experiments to people, but it feels dangerous to artists including me.” When granting copyright to creations of AI, it can put artists in danger. Also, it is impossible to admit the copyright by law as mentioned above, and as the standard of copyright for creations of AI is obscure.
Due to the regulated Copyright Law before the appearance of AI, the pros argued to grant copyrights of AI’s creations. On the other hand, the cons refute with the current law saying only humans are the authors of creative production. These days, AI enormously affects art which was treated as humans’ original fields, as artists utilized AI on their works. Thus, the relative experts should end the discussion by considering effect of AI to humans’ lives and respecting the existing artists.
[1] This platform is usually used to promote works of commercial arts applicants including animations, and games.
[2] It is an AI generating image programs that makes drawings of particular artists’ style when the users put the explanations of pictures and artists’ name.
[3] GitHub Copilot is an AI tool that generates code