연구개발과정에서 인공지능 기술의 조력으로 인한 진보성 요건 기준의 변화- 미국특허법의 비자명성(non-obviousness) 요건을 중심으로 -

The non-obviousness or inventive step requirement is a legal principle established to reward those who achieve innovative technological advancements with the exclusive rights of a patent, recognizing the risky and unpredictable nature of such works. According to Section 103 of the U.S. Patent Act and Article 29(2) of the Korean Patent Act, an invention is not eligible for a patent if it is considered ‘obvious’ or ‘easily-derived’ by a person having ordinary skill in the relevant field (‘PHOSITA’). As we enter an era of paradigm shifts in research and development, facilitated by artificial intelligence (‘AI’) technologies producing outcomes beyond human capabilities, there is a need to reflect the enhanced research and development capabilities brought about by AI in the degree of improvement required by the non-obviousness criterion. In this context, the study discusses the need for a new understanding and adjustment of the concepts of the ‘PHOSITA’ and what is considered ‘obvious’ or ‘easily-derived’. While current AI technologies assist researchers by increasing the efficiency of the R&D process, this study proceeds on the assumption that AI has not yet reached a level where it can replace the creative and inventive roles of human researchers. In both Korea and the United States, the definition of a ‘PHOSITA’ is aligned, referring to a hypothetical person who uses his/her creativity, common sense, background knowledge, and familiar tools to combine prior art teachings like solving a puzzle. This definition was established through the KSR decision by the U.S. Supreme Court. When assessing the skill level of an ordinary technician, if AI technology is commonly and routinely used in that field, the skill level should be adjusted to an ordinary technician who uses AI technology. However, with current technology, the level of outcomes varies depending on the researcher’s AI proficiency. Therefore, the exceptional AI skills of an inventor cannot be equated with the abilities of a ‘PHOSITA’ who uses AI. This study also examines how the criteria for ‘obviousness’ or ‘easily-derived’ should be adjusted due to the active use of AI technology in the R&D process. In the United States, the KSR decision established that an invention is considered obvious if there is a reasonable expectation that a person of ordinary skill in the art could find the solution identified by the inventor at the time of the patent application. Regardless of how advanced AI technology used in the R&D process might be, the assessment of an invention’s non-obviousness considers only the innovation of the research output, not the innovative use of AI technology. Furthermore, as superior search capabilities of AI technology are utilized in the research and development process, all outcomes can be verified and predicted to a certain extent, making them an ‘obvious to try’ technology with a reasonable chance of success. However, according to a 1980s U.S. case law, situations corresponding to the machine learning processes of AI were specifically identified and concluded not to meet the ‘obvious to try’ standard. The examination guidelines of the Korean Intellectual Property Office (KIPO) describe various situations that serve as the basis for determining ‘easily-derived’, but when utilizing AI technology, it has become easy to circumvent these bases for ‘easily-derived’. Currently, the non-obviousness requirement is not particularly difficult for inventions created with the help of AI technology. However, if innovative inventions are pouring out with the help of AI technology, prior arts in this field will quickly become densely populated, leaving no room for further innovation. Or if there arises a need to control the quality and quantity of patents covering such inventions, the current patent system already has the internal capability to adjust the standards for the non-obviousness or the inventive step requirement as needed.