日本における生成AIと法についての現状 - 著作権法を中心に-

Since the introduction of ChatGPT, generative AI has spread rapidly, bringing to light legal challenges such as copyright infringement, deepfakes, and disinformation. In 2025, Japan enacted the AI Act, which established a policy framework including the creation of the AI Strategic Headquarters and basic plans; however, direct regulation of rights infringement remains largely left to soft law measures based on existing legislation. Regarding the Copyright Act, the 2018 amendment introduced “flexible rights limitation provisions” (Articles 30-4, 47-4, and 47-5), which permit the use of data for AI training provided there is no “purpose of enjoying” the thoughts or emotions expressed in the work. In 2024, the Agency for Cultural Affairs issued guidelines on generative AI and copyright, aiming to resolve legal issues within the framework of traditional copyright discussions. These guidelines establish the presence of “enjoyment purpose” as the criterion for the development and training stages, and “reliance” (access) and “similarity” as the criteria for the generation and utilization stages. Regarding the liability of developers and providers of generative AI, the issue of aiding copyright infringement arises. In light of the Winny case precedent, the criteria may include whether generating content that satisfies “reliance” and “similarity” is considered “not exceptional”, and whether the provider recognized and accepted the probability of copyright-infringing use. However, it appears difficult to conclude that the infringing use of generative AI is “not exceptional”. Japan's AI legislation and policies still face many challenges and require continuous deliberation.