1/2024: Artikler
26/03/2024

Introduction and Analysis from legal perspective of Japan’s Latest Draft of AI Guideline(1)This work was supported by JST, ACT-X Grant Number JPMJAX22AA, Japan.

Av Sato Naito, PhD student conducting research on «A» with funding support from ACT-X, a grant program for young researchers provided by the Japan Science and Technology Agency.

1. Introduction

In the EU, the Council and the European Parliament reached a provisional agreement on the AI ACT 2021/0106(COD) in December 2023. While the approaches may differ, many nations share the desire to ensure the safety of AI use, and efforts to establish AI governance policies, such as the AI ACT, are not limited to the EU alone.

Japan influences both the establishment of rules and technological aspects, impacting the EU and the world. Firstly, concerning international guidelines, Japan has led discussions as the chair country during the establishment of the Hiroshima Principles(2)The Government of Japan, The Hiroshima AI Process: Leading the Global Challenge to Shape Inclusive Governance for Generative AI, KIZUNA https://www.japan.go.jp/kizuna/2024/02/hiroshima_ai_process.html , last visited Feb 29th 2024. by the G7. Secondly, in terms of technology, Japanese companies are among those providing AI technology in Europe. For instance, NEC’s subsidiary in Denmark exemplifies its provision of services in the region.

This paper aims to introduce the developments and events in Japan’s AI regulation since the beginning of this year, 2024, including a new guideline draft and interpretation of the Copyright Act, while also highlighting the points that Japan should consider in reference to EU law.

There is limited information regarding Japan’s AI regulations and situation. However, Japan has closely observed regulations and discussions in other countries, such as the EU’s AI ACT. Analyzing and building upon the systems and laws of other countries has been a key factor in Japan’s success in modernization over the past 150 years. Taking a somewhat different approach while considering EU rules and discussions. Japan has adopted a different approach while considering the EU’s regulations and discussions. Japan has its unique background. However, while sharing the same goals, Japan’s policy of taking a different approach may provide valuable insights into certain EU points of contention, such as “regulations like the AI ACT can hinder innovation” and “The AI ACT may have loopholes and may not be effective(3)Veale, Michael and Zuiderveen Borgesius, Frederik, Demystifying the Draft EU Artificial Intelligence Act (July 31, 2021). Computer Law Review International (2021) 22(4) 97-112, Available at SSRN: https://ssrn.com/abstract=3896852”.

Therefore, Japan’s draft might serve as a reference point for future EU discussions and elsewhere discussions. Thus, this paper aims to introduce the developments and events in Japan’s AI regulation since the beginning of this year, 2024, including a new guideline draft and interpretation of the Copyright Act, while also highlighting the points that Japan should consider in reference to EU law.

Illustrasjon: Colourbox.com

2. AI regulation in Japan

In Japan, there is no comprehensive AI regulation through hard law, and there are currently no plans to enact hard law regulations. Unlike the EU, rules in Japan have been established through soft law guidelines on a sector-by-sector basis. Currently, Japan aims to consolidate the three guidelines, and as of 2024, the draft AI Guidelines for Business has been released, which will be discussed further in Next Section 2.1.

2.1 AI Guidelines for Business(4) The provisional English translation can be referenced from the following. Ministry of Internal Affairs and Communications & Ministry of Economy, Trade and Industry, (Draft) AI Guidelines for Business, Jan 2024, https://www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/20240119_4.pdf,last visited Feb 29th2024.

On January 19, 2024, the draft of the AI Guidelines for Business was published. This guideline aims to promote innovation and reduce risks in the use of AI. This section will explain the background of the new guideline, why Japan has guidelines for businesses, adopt a soft law approach, and provide an overview of the new guideline.

2.1.1 The background of the new guideline

There have been three guidelines in Japan regarding AI, “Draft AI R&D GUIDELINES for International Discussions(5)The Conference toward AI Network Society, Draft AI R&D GUIDELINES for International Discussions, Jul 2017, https://www.soumu.go.jp/main_content/000507517.pdf, last visited Feb 29th 2024.”, “AI Utilization Guidelines Practical Reference for AI utilization(6)The Conference toward AI Network Society, AI Utilization Guidelines Practical Reference for AI utilization, Aug 2019, https://www.soumu.go.jp/main_content/000658284.pdf, last visited Feb 29th 2024.”, and “Governance Guidelines for Implementation of AI Principles Ver. 1.1(7)Expert Group on How AI Principles Should be Implemented AI Governance Guidelines WG, Governance Guidelines for Implementation of AI Principles Ver. 1.1, Jan 2022, https://www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/20220128_2.pdf, last visited Feb 29th 2024.”. Japanese Government has decided to integrate and review these three guidelines. Taking into account the characteristics of AI technology that have further developed in recent years and discussions on the social implementation of AI both domestically and internationally(8)Ministry of Internal Affairs and Communications & Ministry of Economy, Trade and Industry,Outline of the draft “AI Guidelines for Business”, page 6, Jan 2024 https://www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/20240119_6.pdf ,last visited Feb 29th 2024., the new guideline is formulated for businesses, including the central and local Government, to practice the social implementation and governance of AI(9)Ministry of Internal Affairs and Communications & Ministry of Economy, Trade and Industry, supra note 3 at page 3..

2.1.2 Why does Japan choose to have guidelines for businesses and adopt a soft law approach?

Compared to the AI Act, the distinguishing features of this new draft guideline are that it applies only to businesses(10)However, it has been suggested during the review meetings of this draft guideline that it would be desirable for non-business entities to also refer to this guideline. Ministry of Economy, Trade and Industry, Summary of Discussions at the Second Meeting of the Ministry of Economy, Trade and Industry’s AI Guidelines for Business, page 1, https://www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/2023_002_gijiyoshi.pdf, only Japanese version is available, last visited Feb 29th 2024. In addition, the guideline itself states that it can be also useful for non-business, including consumers. Ministry of Internal Affairs and Communications & Ministry of Economy, Trade and Industry, supra note 4 at page. and adopts a soft law approach. The approaches of soft law and having guidelines for businesses have not emerged suddenly. Even in domestic regulations, the approaches have been consistently followed since the guidelines of 2019, and this approach continues to be adhered to. Therefore, drawing upon previous documents from the Japanese Government and other sources, this section discusses why Japan adopts these approaches.

According to official documents from the Japanese Government, the adoption of a soft law approach is necessary not only because some businesses desire it but also because there is a desire to establish effective rules. This section explains the rationale behind issuing guidelines targeted at businesses first. Explaining from this perspective elucidates why the adoption of a soft law approach is necessary in Japan.

The reason for targeting businesses is that they are primarily responsible for AI governance. Business activities have become increasingly sophisticated, complex, digitalized, and globalized, making it more challenging for governments to comprehensively understand and monitor them from external perspectives(11)The Ministry of Economy, Trade and Industry, GOVERNANCE INNOVATION Ver. 2: A Guide to Designing and Implementing Agile Governance, page 67, Jul 2021, https://www.meti.go.jp/shingikai/mono_info_service/governance_model_kento/pdf/20210730_2.pdf, last visited Feb 29th 2024..

Ensure incentives for establishing and operating governance systems and compliance programs within organizations function effectively; it is essential for companies to have a clear understanding of what mechanisms should be put in place. On the other hand, due to the high level of variability in circumstances depending on the size and nature of business activities, it is difficult and not advisable to prescribe a one-size-fits-all governance system for all companies. Particularly under agile governance, where there is an expectation of flexible governance that swiftly cycles through the operation and review of goals, relevant rules, and procedures, there is an anticipated challenge in employing traditional approaches such as hard law or judicial review, which have been traditionally emphasized. In such circumstances, it is believed that to realize the rule of law principle, supervisory authorities should be more proactive in communicating the policies underlying the “goals” and derived behavioral norms and procedural guidelines, thereby enhancing accountability(12)The Ministry of Economy, Trade and Industry, supra note 10, at page 75-76..

2.1.3 The overview of the new guideline

The Japanese Government released the draft of the new guideline and an outline for the draft AI Guidelines for Business(13)The Ministry of Economy, Trade and Industry, supra note 7. in January 2024.

From the publicly available information as of the end of February 2024, explicit reasons for why the risk-based approach has been adopted and what influences it could not be found. However, the guideline itself states, “The concept of this ‘risk-based approach’ is widely shared among AI advanced countries.” (14)Ministry of Internal Affairs and Communications & Ministry of Economy, Trade and Industry, supra note 3 at page 3.Additionally, international rules, such as the EU’s AI ACT, that adopt a risk-based approach have been compiled, indicating that the movements of various countries have likely become a significant reason for Japan to partially adopt the risk-based approach.

2.2 Characteristics of the AI Guidelines for Business

In short, the highlights of Japan’s new AI guideline are twofold: first, balancing innovation and safety, and second, prioritizing effectiveness.

Firstly, the guideline features an abstract definition of AI, which addresses the criticism that the initial AI ACT proposal was too narrow in defining AI, thus potentially inadequately regulating its scope. This abstract definition enables flexibility in adapting to the rapidly evolving field of AI.

Secondly, by not targeting specific technologies, the guideline prevents companies involved in AI development or provision from exploiting loopholes. This approach eliminates the need for a perpetual cycle of updating guidelines to regulate emerging technologies and then revising them again to circumvent such regulations.

Furthermore, merely stating abstract rules may leave businesses uncertain about how to proceed. However, the accompanying documentation of Japan’s guidelines provides specific examples of actions taken by companies and failure cases, offering practical guidance for implementation.

Additionally, international rules, such as the EU’s AI ACT, that adopt a risk-based approach have been compiled, indicating that the movements of various countries have likely become a significant reason for Japan to partially adopt the risk-based approach.

2.3 Concerns with the draft of the AI Guidelines for Business

In the draft guideline, issues such as gender discrimination are primarily focused on. Still, it seems there should be more reflection on concerns regarding minorities and an increase in case studies. Furthermore, especially in Japan, it has been pointed out that there is a lack of awareness of rights, as evidenced by the relatively low number of lawsuits per capita(15)Courts in Japan, Court Data Book 2023, only Japanese version is available, https://www.courts.go.jp/vc-files/courts/2023/databook2023/db2023_212.pdf, last visited Feb 29th 2024.(16)Courts in Japan, STATISTICAL TABLES, https://www.courts.go.jp/english/vc-files/courts-en/file/2022_STATISTICAL_TABLES.pdf, last visited Feb 29th 2024.. Therefore, it is essential to provide robust support, including non-legal means, for safeguarding and preventing rights violations, not only for the government and private sector but also for ensuring that rights infringements do not occur in the first place.

This approach eliminates the need for a perpetual cycle of updating guidelines to regulate emerging technologies and then revising them again to circumvent such regulations.

2.3.1 Consideration of the Nature Unique to the Public Sector

Citizens practically cannot refuse the utilization of AI by the public sector; it tends to become obligatory in a sense. The competitive principle of citizens choosing the more reliable institution also likely does not work at all. Additionally, since AI usage methods targeting almost all citizens of a nation are possible, the number of people affected could be significant. Therefore, there is a higher possibility of widespread rights infringements than private companies. Especially in Japan, there is a lack of awareness regarding legal rights, as seen in the relatively low number of lawsuits per capita compared to other East Asian countries. It is advisable to provide thorough measures, including methods other than litigation, to safeguard and prevent rights violations, regardless of whether they occur in the public or private sector. Therefore, it is necessary to reconsider the relationship between the rights of citizens and public sectors, such as administration, and legislative amendments, such as safeguards, which are required.

In particular, there are issues unrelated to AI concerning biases and profiling in Japan’s public sector. Even the US embassy, where profiling by AI is a global concern in some states, has called attention to profiling by the Japanese police.

This is not a problem of profiling using AI systems but rather a problem of profiling done by humans. However, the issue of bias in AI usage arises not from discrimination within AI systems but from amplifying societal discrimination. While the Japanese Government denies racial profiling, there are already instances of profiling against foreigners pointed out by foreign embassies, implying the need for deeper consideration, even in the context of crime investigation and categories like race or foreigners, to address concerns regarding transparency and fairness. Japan is a signatory to the Convention on the Elimination of All Forms of Racial Discrimination and prohibits racial discrimination in Article 14 of its Constitution. Furthermore, among the rights of citizens stipulated in the Japanese Constitution, one-third pertains to criminal procedures. In the historical context and provisions of criminal procedures, there are repeated mentions of “any person” rather than specifically “Japanese citizens.” From these occurrences, it seems that profiling in investigations and trials should be specifically considered, and it would be desirable to specify such aspects in existing hard law such as the Code of Criminal Procedure. Ironically, the example of COMPAS in the United States demonstrates the realistic dangers of racial profiling in this field and should be heeded by Japan.

2.3.2 Matters to be Further Addressed in Each Item

In the guideline, there are references to human rights, as mentioned above, and discrimination against women is repeatedly addressed.

From the perspective of legal scholars, apart from the issues explicitly stated in the appendix of the guideline, there are specific matters regarding discrimination and bias that need to be addressed. The guideline mentions several times considering the attributes enumerated in Article 14(1) of the Constitution and the rules concerning international human rights(17)See for example, Ministry of Internal Affairs and Communications & Ministry of Economy, Trade and Industry, (Draft) AI Guidelines for Business Appendix, page 78, https://www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/20240119_5.pdf, last visited Feb 29th 2024.. Regarding the Constitution and international human rights rules, there are two points of concern.

Firstly, the current Constitution of Japan, enacted in 1947, remains unamended, making it one of the oldest constitutions in the world. One reason for its lack of amendment is that the Japanese Constitution contains concise and abstract provisions(18)Yokodaido, S. (2019) ‘Constitutional stability in Japan not due to popular approval’,German Law Journal, 20(2), pp. 263–283, allowing for flexible interpretation, which has been carried out by laws enacted by the Japanese parliament and Government. Article 14, paragraph 1 of the Constitution referred to in the guideline states, “All of the people are equal under the law, and there shall be no discrimination in political, economic or social relations because of race, creed, sex, social status or family origin.” However, it is difficult to interpret what exactly is encompassed by “social status.” Moreover, some cases concerning discrimination against women may not be straightforwardly identifiable.

2.3.2.1 Regarding Children

During the process of creating the guideline, it was pointed out that there was no mention of minors. Currently, in the proposed version, children have been included as stakeholders. In Japan, there is a tendency to focus on children›s education, which has led to the increasing use of Edtech provided by private companies. Furthermore, in some administrative processes, AI has already been used for risk assessment in child abuse cases(19)In Tsu City, Mie Prefecture, there was a case where child abuse was overlooked, resulting in the death of a child. Some media outlets raised concerns about the use of AI in this case. However, it cannot be solely attributed to AI, considering that there were previous cases of death before AI was introduced in the municipality, and the fact that the responsible personnel had not confirmed the child’s situation for about a year in this particular case., sparking discussions.

Children have limited means to collect accurate information, express their opinions, and have them reflected compared to adults. Therefore, more should be done beyond just including them as stakeholders. Consider the nature of children’s rights and the dissemination of information regarding AI, drawing lessons from discussions within GDPR and AIA of the EU and its member states. For example, there have been references to providing information in a way that is understandable to children in Finland (20)Eduskunta, Regeringens proposition RP 145/2022 rd. Senast publicerat 20-09-2022, https://www.eduskunta.fi/SV/vaski/HallituksenEsitys/Sidor/RP_145+2022.aspx,, and such approaches could serve as references for Japan.

2.3.2.2 Regarding people with disabilities

Japan has ratified the Convention on the Rights of Persons with Disabilities. However, within the scope of my search, it seems that neither the main body of these guidelines nor any attached documents explicitly mention people with disabilities. In EU member states, for example, considerations have been made regarding individuals with disabilities in the context of legislative amendments concerning automated decision-making by the administration(21)For example, there are references to the environment surrounding individuals with disabilities, such as the existence of dispute resolution committees for individuals with disabilities, in Finland. Eduskunta, Regeringens proposition RP 145/2022 rd. Senast publicerat 20-09-2022. https://www.eduskunta.fi/SV/vaski/HallituksenEsitys/Sidor/RP_145+2022.aspx, and amendments related to automation have been considered with this in mind. To prevent unjust discrimination and bias, it is necessary to provide concrete examples of individuals with disabilities or to understand and evaluate the situations surrounding minorities, such as individuals with disabilities.

One of the reasons Japan leaned heavily towards AI development, compared to the EU, was because Japan lagged behind the United States and China in AI development and adopted “progressive” policies as desired.

3. Latest events in the Inter­pretation Change of Copyright Law Regarding AI in Japan

As supplementary information, I would like to introduce the latest discussion in Japan regarding copyright law, particularly the interpretation proposal of the new copyright law presented by the Government at the end of February(22) Materials distributed at the subcommittee can be downloaded from the following link. However, as of the end of February 2024, all of the materials are only available in Japanese. Agency for Cultural Affairs, Subcommittee on Copyright Legislation of the Cultural Affairs Council (7th Session), Feb 2024,https://www.bunka.go.jp/seisaku/bunkashingikai/chosakuken/hoseido/r05_07/,last visited Mar 2nd2024.. In essence, the scope of using copyrighted materials for AI learning without the copyright holder’s consent has been narrowed compared to previous interpretations.

Under the Copyright Act, copyrighted materials could be used for AI learning and development without the copyright holder’s consent as long as they do not unfairly harm the copyright holder’s rights. Therefore, Japan has been considered a “machine learning paradise” (23)Tatsuhiro UENO, Copyright Issues on Artificial Intelligence and Machine Learning, https://www.f.waseda.jp/uenot/Copyright-AI-IJCAI2017.pdf, last visited Mar 2nd 2024. at least since the 2010s. However, the new interpretation now considers, for example, generating works from a small amount of copyrighted materials from specific creators as copyright infringement(24)Agency for Cultural Affairs, Perspectives on AI and Copyright (Draft), Feb 2024,https://www.bunka.go.jp/seisaku/bunkashingikai/chosakuken/hoseido/r05_07/pdf/94011401_03.pdf , last visited Mar 2nd 2024..

One of the reasons Japan leaned heavily towards AI development, compared to the EU, was because Japan lagged behind the United States and China in AI development and adopted “progressive” policies as desired. For example, Japan allowed the free downloading of data for AI systems, earning it the nickname “heaven.” Naturally, there have been movements opposing such policies from creators recently. A group called the “Group of volunteers who concern the future of creators and AI” was formed and made proposals in public(25)Group of volunteers who concerns future of creators and AI, https://support-creators.com/start_page, last visited Mar 2nd 2024. While the second edition is available only in Japanese, the first edition can be referenced in English. in 2023.

It is meaningless to regulate downloads in other countries if one country allows downloads and all AI development without the consent of copyright holders. Considering that provisions related to copyright are ineffective if they do not align transnationally, Japan’s policy revision in this regard seems to be welcome.

[…] Japan’s approach may address concerns such as loopholes in laws mentioned in the EU’s AI ACT and the inhibition of innovation by hard law.

4. Conclusion

The guidelines in Japan are still in the draft stage, and it’s unclear what their final form will be. Since the primary objective of Japan’s guidelines is effectiveness, evaluating whether they truly achieve effectiveness will require time and results. Ultimately, whether Japan’s guidelines become significantly meaningful from a global perspective regarding effectiveness remains uncertain. Additionally, it’s worth noting that Japan has a culture of adhering to government policies even without legal binding, as evidenced by its ability to provide AI technology domestically and its response to COVID-19 with minimal use of hard law. However, it cannot be denied that some aspects unique to Japan contribute to this, as mentioned.

Nevertheless, Japan’s approach may address concerns such as loopholes in laws mentioned in the EU’s AI ACT and the inhibition of innovation by hard law. Therefore, there is potential for Japan’s approach to respond to these issues effectively. On the contrary, there is much that EU law and the laws within EU member states can teach Japan, such as the protection of minority rights. Since they share similar overarching goals, comparing the rules to be implemented in each country will become increasingly crucial for better rule-making.

Sato Naito