1. Positive aspects of AI systems

The emergence of artificial intelligence (AI) systems, coupled with the rapid advancement and widespread adoption of this technology, has made our lives significantly more convenient. This trend is expected to continue in the future.

Even in my own daily life, the computers I use are equipped with AI systems. I engage in casual conversations with AI, use it to draft more natural-sounding messages when corresponding by email with researchers overseas, and exchange ideas through a kind of back-and-forth dialogue to develop research concepts. Such experiences would have been completely unimaginable only a few years ago, and these technological innovations have already provided considerable benefits to my own research activities. (I would like to emphasize that I did not simply delegate the preparation of this article entirely to AI.) More broadly, in my everyday life, before writing this manuscript, I used a smartphone application to operate a robotic vacuum cleaner, which thoroughly cleaned the floors in the living and dining areas of my home. This has greatly reduced the time required for household chores and has made a significant contribution to creating the time needed to write this article.

Shifting our perspective from individual experiences to society as a whole, technological innovations in AI systems are likely to make autonomous driving a reality in the not-too-distant future, eventually ushering in an era of fully autonomous vehicles. From a logistics perspective, this could help address labor shortages in the transportation sector. Moreover, from the standpoint of personal preference, tourism, and travel, time that was previously spent operating a vehicle could instead be used to enjoy more meaningful moments with family and friends.

2. Negative aspects of AI systems

While technological innovations represented by AI systems may appear to offer significant benefits, we must also recognize that, as with any new technology, they may give rise to unforeseen accidents. When victims suffer damages as a result of accidents arising from the negative aspects of technological innovation, they may seek civil liability against the responsible parties based on the Civil Code or the Product Liability Act. In particular, when an accident is caused by a product, the Product Liability Act is expected to play a central role.

However, using AI systems as an example, it may be difficult to determine the reasons behind a system's particular actions. Moreover, when an AI system is incorporated into a product, the system may, through self-learning, behave in ways that the manufacturer did not anticipate or cannot adequately control. If a product is connected to the internet, there is also the risk that a cyberattack could cause the product to malfunction or behave in an unintended manner. Furthermore, when software is incorporated into a product and an accident occurs due to a defect in that software, the issue may be addressed under the current legal framework by considering the product itself to be defective. However, as technological innovation progresses, software may no longer be incorporated into a product but may instead exist independently online. If such software contains a defect and causes a similar accident, a fundamental question of legal interpretation arises as to whether the software itself can be considered a "product" under the Product Liability Act. Thus, technological innovations represented by AI systems are likely to create risks and types of accidents that could not previously have been anticipated. To address these emerging risks, the Product Liability Act will need to respond through both the development of legal interpretation and further consideration of legislative measures.

3. The possibility of product liability legislation for AI systems: Insights from EU law

The Ministry of Economy, Trade and Industry recently published the Guidance on the Interpretation and Application of Civil Liability in the Utilization and Application of AI [Version 1.0]. This is an important document that may serve as a catalyst for future developments in legal interpretation and legislative discussions.

Turning our attention overseas, the European Union (EU) has adopted an approach that seeks to strike a balance between regulating technological innovations represented by AI systems and promoting their continued development and widespread adoption. In 2024, the Product Liability Directive (Directive (EU) 2024/2853) entered into force, requiring Member States to transpose the Directive into their national laws by the end of 2026. Accordingly, EU countries are currently undertaking revisions to their product liability legislation. The Directive expressly includes software within the definition of a product. It also incorporates AI-related factors, including self-learning capabilities, when determining whether a product is defective, and addresses technological innovations by taking cybersecurity-related factors into account in anticipation of internet-connected products. Furthermore, it allows the relevant point in time for assessing defects to be set later than the time of circulation in certain circumstances, in order to accommodate software updates and upgrades implemented to address cybersecurity issues. In addition, the Directive expands the scope of liable parties in response to increasingly diverse distribution channels, establishes rules regarding the disclosure of evidence, reduces the burden of proving defects and causation, and eliminates provisions that cap liability amounts. All of these features can be evaluated as intending to modernize product liability law in response to technological innovations represented by AI systems.

With regard to the Product Liability Directive, one particularly noteworthy aspect is that it establishes rules allowing public-law regulations concerning AI systems, such as the AI Act, to be taken into account when determining whether a product is defective and when alleviating the burden of proof. Through the Product Liability Directive, the EU appears to have adopted an approach under which the prior regulation of AI systems is addressed through public regulation, while subsequent issues are handled through civil liability and product liability, which serve as the function of providing compensation for damages. In other words, the EU seeks to establish a coordinated and consistent framework through cooperation between public law and private law. Understanding the EU's approach in this manner suggests that it is premised on the possibility of regulating AI systems. On this basis, the EU can be viewed as having presented an approach to addressing potential civil liability arising from accidents caused by AI systems.

4. How Japanese law should handle insights from EU law

Around the time of the revision of the Product Liability Directive, the EU sought to take a global leadership role in addressing technological innovations represented by AI systems by introducing various laws and regulations concerning AI, including the AI Act. This approach may also be viewed as an effort to pursue what is known as the Brussels Effect, a term derived from the fact that the EU's headquarters is located in Brussels, the capital of Belgium. At the same time, however, it should not be overlooked that the approach adopted toward AI systems is closely linked to the economic policies of individual countries. If Japan were to refrain from adopting the EU's approach of imposing strict regulations on AI systems and instead pursue a laxer regulatory stance, it would be necessary to avoid directly importing the EU Product Liability Directive into Japanese law and revising the Product Liability Act accordingly.

On the other hand, both before and after the revision process for the EU Product Liability Directive, a vast number of academic papers and reports have been published. Indeed, the volume of research is so extensive that one might even wish to translate and analyze these materials with the assistance of AI systems. These works provide not only detailed discussions of the interpretation of individual provisions of the Product Liability Directive, but also multidimensional perspectives, including alternative forms of liability outside the product liability framework, such as strict liability and risk liability; the possibility of granting legal personality to AI systems; and even the establishment of compensation fund schemes separate from the imposition of liability itself. This body of research can be regarded as a treasure trove of ideas, offering abundant insights not only into the appropriate form of public-law regulation of AI systems and its relationship with determinations of civil liability, but also into the design of civil liability schemes for accidents caused by AI systems and mechanisms to ensure their effectiveness. By drawing upon this wealth of research and analyzing and examining the available materials, it may be possible to gain insights into the development of a Japanese framework for civil liability arising from accidents caused by AI systems. I intend to continue my research in this area as well.

---

[Reference]

• Otsuka, T. "Genkōhōka de no AI ni kansuru fuhōkōi sekinin ni tsuite" in The Future Vision of Civil Law and Civil Law Scholarship: A Collection of Essays in Memory of Professor Atsumi Kubota, (Shojihomu, 2026), p. 261.
• Shiraishi, T. "AI・avatā no fuhōkōi sekinin" (Zeimu Keiri Kyokai, 2026).
• METI, Guidance on the Interpretation and Application of Civil Liability in the Utilization and Application of AI [Version 1.0].

Taro Maeda／Professor, Chuo Law School
Area of Specialization: Civil Law

Taro Maeda was born in Kanagawa Prefecture in 1981. He graduated from the School of Law, Waseda University in 2004. Afterwards, he completed the Master’s Program in the Graduate School of Law, Waseda University in 2007, and then completed the Doctoral Program without obtaining a degree in the same graduate school in 2012. He then served as a Research Associate in the Faculty of Law, Waseda University, a Full-Time Lecturer in the Faculty of Law, Aichi Gakuin University, and an Associate Professor in the Faculty of Law, Aichi Gakuin University. He was appointed as Associate Professor in the Chuo Law School in 2020. He has held his current position since 2023.

His current research focuses on the interpretation and legislative development of civil liability for accidents arising from AI systems, with comparative studies of EU law, German law, and Austrian law.

His recent major publications include “Minpō 714-jō ni motozuku oya no fuhōkōi sekinin no kiseki kōzō no saikōsei” in Research on Tort Law: Vol. 6, edited by Ikufumi Niimi, Michitaro Urakawa, and Hanae Furuya, (Seibundo, 2025), from p. 39 onward, “The Significance of the Plurality of Liability Principles in Tort Law and Examination of Their Interrelationships (1)–(3, Complete): Insights from Austrian and German Laws”, Chuo Law Journal, Vol. 18, No. 3, from p. 53 onward; Vol. 18, No. 4, from p. 99 onward; and Vol. 19, No. 1, from p. 65 onward (2021–2022), and more.

• Share articles
• 
• 
•