AI Regulation in Japan
AI Promotion Act
On 4 June 2025, Japan enacted the Act on Promotion of Research and Development, and Utilization of AI‑related Technology (the “AI Promotion Act”). This is Japan’s first piece of AI‑specific legislation. Although the AI Promotion Act does not explicitly address its applicability to foreign entities, discussions in the Diet (the national legislature of Japan) clarified that it is expected to apply to them as well. Therefore, foreign entities intending to engage in AI‑related business in Japan should take the AI Promotion Act into consideration.
The AI Promotion Act primarily sets out basic policies for AI research, development, and utilization. It provides a framework for the government’s formulation of AI‑related basic plans, the implementation of basic policies, and the establishment of a national headquarters for AI strategy. Many provisions function as foundational guidelines directed at national and local governments. In addition, the AI Promotion Act anticipates that guidelines will be established to ensure the appropriate research, development, and use of AI‑related technologies in alignment with international norms.
The obligations imposed on private enterprises, set forth in Article 7, are limited. This provision requires any business entity that utilizes AI‑related technologies (an “AI-utilizing business entity”) to co-operate with measures implemented by national and local governments under the AI Promotion Act. Article 7 also encourages AI-utilizing business entities to improve the efficiency and sophistication of their business activities and to foster the creation of new industries through proactive use of AI‑related technologies.
Although the AI Promotion Act does not prescribe penalties, Article 16 authorizes the national government to issue administrative guidance and advice to an AI-utilizing business entity that violates its obligations thereunder. The government may also publicly disclose relevant information, including – when deemed necessary – the name of the violating AI-utilizing business entity.
In accordance with the AI Promotion Act, the government approved the “AI Basic Plan” on 23 December 2025, and the “Guidelines for Ensuring Appropriateness of Research, Development and Utilization of AI Related Technology” (the “Guidelines for Appropriateness”) on 19 December 2025. Under the Guidelines for Appropriateness, AI utilizing business entities are expected to adhere to the following principles:
-
ensuring holistic appropriateness through AI governance;
-
ensuring transparency to build trust with stakeholders;
-
ensuring adequate safety;
-
maintaining a safe environment by securing business continuity; and
-
considering stakeholders in light of the importance of data as the foundation for AI innovation.
Although the Guidelines for Appropriateness are not legally binding, they are intended to serve as a reference for anyone conducting AI related business in Japan.
AI Guidelines for Business
Because the AI Promotion Act sets out basic policies, despite its enactment, soft law continues to be important for AI-related initiatives in Japan. Among these, the AI Guidelines for Business (the “AI Guidelines”) are frequently referenced when companies utilize AI.
The AI Guidelines provide AI developers, providers and users (collectively “AI Business Operators”) with fundamental principles regarding AI-related risks and countermeasures. Version 1.0 of the AI Guidelines was formulated in April 2024, but the AI Guidelines continue to be updated, and Version 1.2 was published in March 2026. Among other things, the AI Guidelines highlight the importance of AI governance. They provide that, in order to respond to complex and rapid changes, it is important to practice “agile governance,” in which a cycle of “environmental and risk analysis,” “goal setting,” “system design,” “operations,” and “evaluation” is continuously and rapidly iterated among multiple stakeholders.
Civil Liabilities of Utilizing AI
Legal Actions Regarding Copyright Infringement
In August 2025, The Yomiuri Shimbun, one of Japan’s leading newspaper companies, filed suit against Perplexity in the Tokyo District Court, seeking an injunction against alleged acts of copyright infringement and approximately JPY2.1 billion in damages. Subsequently, Nikkei Inc. and The Asahi Shimbun Company filed similar lawsuits. These publishers allege that Perplexity infringed their copyrights by using their news articles to train its AI and by providing summaries of these articles without authorization. In addition, 48 companies affiliated with Kyodo News issued a warning letter demanding that Perplexity cease the unauthorized use of their articles and compensate them for damages, asserting that such use constitutes copyright infringement.
Although the courts have not yet issued decisions in these cases, any ruling will likely become the first judicial precedent in Japan concerning the use of copyrighted works in the development of generative AI. Accordingly, these proceedings warrant close attention.
Civil Liabilities in Utilizing AI
Though the AI Promotion Act is in full force and effect, it does not directly regulate AI business engaged by AI Business Operators. Accordingly, liability for damages arising from AI services will be determined within the existing legal framework. Nevertheless, there are virtually no existing laws that contemplated the existence of AI at the time of their enactment, and there are not enough court precedents to adequately grasp the court’s tendencies with respect to, or unified judicial interpretation that takes into account, the black-box nature and autonomy of AI. In the current environment—where advances and expansion in AI technology are remarkable, including agentic AI and physical AI, and where the likelihood is growing that AI decision-making outputs could directly cause infringement of rights, interests, or harm to life and bodily integrity in the real world—it is not clear under what circumstances AI Business Operators may be liable or what kind of liability they may incur.
In light of the foregoing, on April 9, 2026, the Ministry of Economy, Trade and Industry (METI) published the “Guide on the Interpretation and Application of Civil Liability in AI Utilization [Version 1.0]” (the “METI Guide“), which organizes the concept of civil liability in the utilization of AI based on existing laws and court precedents. The METI Guide examines tort liability in cases where damage is caused to third parties in connection with the utilization of AI. The METI Guide classifies AI into two categories based on the manner of use—”Assistive/Supportive AI” and “Reliance/Substitutive AI”—and examines the direction of civil liability determinations and liability in hypothetical cases for each category.
<Assistive/Supportive AI>
Assistive/Supportive AI refers to a type of AI that is used solely to assist or support the judgment of AI users, and is designed so that human judgment and action are ultimately involved. Many AI systems currently in use likely fall under this category.
In the METI Guide, when AI falls under the category of Assistive/Supportive AI, determinations of civil liability for AI Business Operators are made on the basis of the following, pursuant to existing laws and court precedents.
|
Categories of AI Business Operators
|
Determinations of Civil Liability
|
|
(i) AI Users
|
The standard of the duty of care is not affected by whether or not AI is used, and appropriate judgment and action must be taken under the individual circumstances.
|
|
(ii) AI Developers and Providers
|
While it is assumed that AI users will judge the appropriateness of outputs from AI, explanations regarding limitation of performance and significant risks of AI, as well as certain design measures addressing risks that are not easily foreseeable or manageable by AI users, may be required.
|
<Reliance/Substitutive AI>
Reliance/Substitutive AI refers to a type of AI that, unlike Assistive/Supportive AI, is intended to function as a substitute for human judgment and action, and is to be used in reliance on the AI’s output. AI falls into this category if the following are met: (1) necessity: there must be an expected utility that would be difficult to achieve through the involvement of human judgment and action; and (2) accuracy and safety: the AI must possess a certain level of accuracy and safety.
In the METI Guide, when AI falls under the category of Reliance/Substitutive AI, determinations of civil liability for AI Business Operators are made on the basis of the following, pursuant to existing laws and court precedents.
|
Categories of AI Business Operators
|
Determinations of Civil Liability
|
|
(i) AI Users
|
The duty of care shifts from the general tort liability standard requiring individuals to exercise reasonable judgment and action, to a duty to properly design business processes incorporating AI systems and to operate them while reducing risks as much as possible.
|
|
(ii) AI Developers and Providers
|
Reasonably feasible design measures to ensure and maintain the safety of Reliance/Substitutive AI are required, as well as explanatory measures such as analyzing information important for risk control and providing that information to AI users.
|
The METI Guide states that compliance or non-compliance with the AI Guidelines does not immediately determine the presence or absence of negligence. It also provides, however, that, in light of the framework for assessing negligence, (i) where a party has conducted risk investigation and analysis, and established systems for such investigation and analysis, taking into account the principles set forth in the AI Guidelines, and a unique risk that was nonetheless difficult to foresee has materialized, this may be taken into consideration as a factor weighing against a finding of foreseeability and (ii) where a party has implemented practically reasonable countermeasures commensurate with the degree of anticipated risk, this may be taken into consideration as a factor reducing the likelihood of being found in breach of the duty to avoid harm. The fact that the METI Guide suggests that the establishment and operation of appropriate AI governance may affect civil liability in the event of future damage is quite important, as it provides AI Business Operators with an incentive to establish and operate AI governance.
It should be noted that the METI Guide was prepared by a government agency and is not binding on the courts. Nevertheless, because it was compiled by experts based on existing case law, it is a valuable reference when examining civil liability in connection with the utilization of AI.
Conclusion
In Japan, the AI Promotion Act has been enacted, but soft law such as government guidelines continues to play an important role with respect to AI Business Operators’ utilization of AI. The Japanese government has published various guidelines and guidance documents within a short period of time, and it is necessary to closely monitor these developments in AI practice.
