In March 2026, in a case concerning whether an invention allegedly generated autonomously by AI could be recognized as patentable, the Supreme Court of Japan decided not to accept the applicant’s final appeal, thereby effectively upholding the lower courts’ rulings that “an inventor is limited to a human being.” As a result, it is now judicially settled that, under current Japanese law, an AI itself cannot be named as an inventor in a patent application. In the application at issue, the inventor field identified the inventor as “DABUS, an artificial intelligence that autonomously invented the present invention.” Although the Japan Patent Office ordered the applicant to amend the inventor designation to a natural person, the application was ultimately dismissed because the applicant refused to comply.
At first glance, this decision may appear to reflect a harsh conclusion in the age of AI. However, the core of the ruling is not about “how AI technology should be evaluated,” but rather about “how the wording and structure of the current legal framework should be interpreted.” The Tokyo District Court reasoned that, in light of Article 2, Paragraph 1 of the Basic Act on Intellectual Property, which defines intellectual property as something “created through human creative activities,” it is appropriate to understand an invention as something created by a natural person. Furthermore, in patent application practice, the Japan Patent Office expressly states that the inventor indicated in an application “must be identified by the name and address or domicile of a natural person (individual).” In other words, it is more accurate to understand the court’s conclusion not as a policy-driven exclusion of AI, but as a straightforward confirmation of the premises underlying the current system.
What is important here is the need to distinguish between “an invention involving AI” and “an invention for which only AI is named as the inventor.” What was rejected in this case was, strictly speaking, the permissibility of an application that identifies AI alone as the inventor. Put differently, this does not mean that all inventions in which a human used AI as a tool and contributed creatively in the process have been categorically denied. In fact, the United States Patent and Trademark Office has indicated, through its guidance, that while inventors must be natural persons, AI-assisted inventions are not automatically excluded, and a human may qualify as an inventor if that person made a significant contribution to the conception of the invention. The European Patent Office has likewise held in the DABUS case that an inventor must be a natural person. Even from an international perspective, the prevailing approach today is that “AI is not an inventor, but inventions made using AI may be protected depending on the extent of human involvement.”
In that sense, the Supreme Court’s decision is better understood not as a ruling that stifles Japanese innovation, but rather as one that brings institutional challenges into clearer view. The Intellectual Property High Court’s judgment also strongly reflects the view that the proper protection of AI-generated inventions is a matter that should be addressed legislatively, rather than resolved solely through interpretation of the current law. The Japan Patent Office, too, has identified as issues for consideration such matters as how the patent system should respond to advances in AI technology, false claims of inventorship, and whether AI-generated inventions are eligible to serve as cited inventions. The issue is no longer one that can be exhausted simply by asking whether “AI can be an inventor.” Questions such as who should be named as the inventor, how AI-generated technical ideas should be incorporated into the disclosure system, and how to prevent applications filed under incorrectly named human inventors are likely to become central to future institutional debates.
From a practical standpoint, the lessons that companies and R&D divisions should take from this news are clear. First, at least under current Japanese law, AI cannot be listed as such in the inventor field of a patent application. Second, where results are generated using AI, it will be necessary to document, more carefully than ever before, who was creatively involved, at what stage, and in what manner. Unless one properly organizes where the human made a substantive contribution—whether in designing prompts, defining the problem, selecting outputs, improving them, or verifying them through experimentation—disputes may later arise concerning inventorship or ownership of rights. This decision does not put the brakes on the use of AI itself; rather, it shows that the key dividing line in future intellectual property strategy will be whether human involvement can be adequately explained.
In my view, the decision itself is quite sound under the current law, and from the court’s standpoint it was a natural conclusion. At the same time, social reality has already begun to outpace the assumptions built into the existing legal framework. Now that AI is no longer merely an auxiliary tool, but a technology capable of presenting large numbers of technical options rapidly and at scale, a purely formalistic rule that “inventors must be human” is likely to lose explanatory power in a growing number of situations in the near future. That is precisely why what is needed is neither an attempt to stretch the law through forced interpretation nor, conversely, a blanket exclusion of AI-generated outputs. What is needed is to redesign the concepts of inventorship, ownership of rights, and disclosure in a way that reflects the actual nature of creation in the AI era.
If read only as establishing the conclusion that “AI has no qualification to be an inventor,” this Supreme Court decision may appear to be conservative news. But its true significance lies beyond that. Japanese patent law is now being asked to take the next step toward adapting to the age of AI. Should this decision not be regarded as the starting point for that discussion?