Japan’s AI regulation for businesses

AI’s development is accelerating exponentially. While its regulation is in its infancy, one dramatic innovation seems to lead to the next at a staggering pace. Generative AI technology that creates art, literature, journalism reports, translations, even academic papers, is but one astonishing example.

Historically, the rise of seminal technologies births social, political and legal issues that require due care to assure the novel technology is a net positive for society. AI is perhaps humanity’s most impactful invention ever, and while businesses are already reaping the benefits of intelligent automation, it’s not hyperbole that AI presents a myriad of civilisational defining issues – human rights, ethics, privacy, bias, labour rights, etc.

There has been no global regulatory framework for AI, and so far liberal democracies are employing differing approaches aimed at balancing risks against benefits. A common concern is that over-regulation will harm innovation.

Even so, in December 2023, the EU opted for a comprehensive binding regulation with its AI Act, a scheme targeting specific digital environments that will be overseen by the European AI Office. Britain has chosen a lighter pro-business approach, dividing governance between existing regulators. The US is evolving along UK lines, although there is a call for a federal AI Bill of Rights and a recent presidential executive order instructing federal agencies to formulate AI regulation plans.

Japan is developing its own approach to AI regulation. While a civil law system by tradition, Japan tracks the common law UK and US methods in some legal fields, with government ministries issuing guidelines and existing laws governing the new legal issues of AI.

INDUSTRY GUIDELINES

In January 2024, the Ministry of Internal Affairs and Communications and the Ministry of Economy, Trade and Industry issued the Draft AI Guidelines for Businesses, updating and integrating previous principles and guidelines followed by the compilation of the Hiroshima AI Process Comprehensive Policy Framework at the G7 Hiroshima Summit in May 2023.

These guidelines cover developers, providers and users of AI in business and government. Users for non-business activities are excluded. So are data providers, since the responsibility for handling data should be assumed by the developers and providers.

The guidelines lay out three fundamental values in an AI society: human dignity; diversity and inclusiveness; and sustainable society. They are centred around 10 common guiding principles: human-centric, safety, fairness, privacy protection, security, transparency, accountability, education/literacy, fair competition and innovation. The guidelines call for AI developers, providers and users for business purposes to be mindful of people and the environment in line with the common guiding principles, by way of:

• Combating bias in the development and use of AI, via data collection and management procedures and algorithmic measures;
• Security measures, e.g. ensuring verifiability during AI development;
• Protection of privacy, e.g. inappropriate input, personal information, etc.; and
• Agile governance responsive to environment/risk analysis, goal setting, system design, operation and evaluation continually and rapidly rotated by multiple stakeholders, rather than fixed rules and procedures. The draft guidelines provide governance action items, and its appendix offers practical examples.

Business actors in advanced AI are advised to follow the Hiroshima Process International Guiding Principles for All AI Actors and the Hiroshima Process International Code of Conduct for Organisations Developing Advanced AI Systems. Developers of advanced AI systems should also comply with the International Code of Conduct for Hiroshima Process for Organisations Developing Advanced AI Systems.

Sector-specific guidelines have also been published such as the use of generative AI in education and by healthcare providers. While regulations are non-binding, actors in AI in Japan are expected to know and follow these guidelines and other relevant guidelines.

EXISTING LAWS FOR AI ISSUES

In addition to the guidelines, Japan has been governing AI with existing laws rather than industry-specific legislation. The development, deployment and use of AI, like any other business activity, are subject to, among others, the Copyright Act, the Personal Information Protection Law, the Unfair Competition Prevention Act, the Antimonopoly Law, the Economic Security Promotion Act and, possibly, the Lawyers Act. Here are a few examples:

Copyright Law. AI raises issues on infringement in learning/development, infringement in generation and use, and copyrightability of products produced by AI. For example, building AI knowledge exploits information which may be subject to copyright.

In principle, the use of works requires the permission of the copyright holder. The established narrow exception is if the purpose is not “for enjoying oneself or having others enjoy the ideas or emotions expressed in the works”, such use does not require any permission. However, if the user has a purpose of enjoyment (e.g. overfitting), the user may be subject to a copyright infringement claim including financial damages and injunctive relief (e.g. the disposal of learned models).

Infringement during the generation and use of AI is determined using the conventional standard for infringement (similarity and reliance with existing works) as if the work were created by a human being without using AI. Infringement in the generation and use phase may result in liability not only for the AI user but also for the provider of the generative AI. Copyrightability of AI-generated work follows the traditional interpretation of copyrighted work. If the instructions given to the generated AI are merely ideas that do not lead to expression, no creativity is recognised and the work is not copyrightable.

Personal Information Protection Law. The use of personal information with AI is governed by the existing Personal Information Protection Law. When providers of generative AI (e.g. OpenAI, which provides ChatGPT) intend to capture personal information to develop generative AI, they should ensure that they have obtained personal information by correctly indicating the purpose of use, that the actual method of use does not deviate from the purpose of use, and that they do not violate rules for obtaining sensitive information such as race and creed. In June 2023, an alert was issued to OpenAI requesting notification of the purpose of using personal information in Japanese, and proper acquisition of sensitive information.

The user of generative AI also needs to be careful about personal information. For example, a company that manages personal information of its employees may input personal information when using generative AI to improve operational efficiency. In this case, care should be taken to ensure that the law is not violated.

If the personal data is used to teach generative AI, the party inputting the data may be construed to have provided personal information to a third party (or in the case of using foreign generative AI, to have transferred personal information to a foreign country) in violation of the law.

Competition Law. The government is scrutinising AI for violations of the Antimonopoly Law. In March 2021, a report titled Algorithms/AI and Competition Policy, sponsored by the Japan Fair Trade Commission, was published. It states that while pricing and price research algorithms may increase price competition, depending on how they are used, competitors may use AI to co-ordinate price adjustments, interfere with transactions between competing users, consumer predatory pricing, intentional ranking manipulation, etc.

Economic Security Promotion Act. In May 2022, AI was designated as a “specific critical technology” (including its use by foreign parties to undermine national security). As a result, government support is available to promote the research and development of AI. At the same time, measures must be taken against leaks of technology overseas, protection of trade secrets, and ensuring sound and fair research.

The regulatory scheme for AI in Japan may appear piecemeal or otherwise insufficient to some, but this light-touch approach of thoughtful government and industry guidelines, and application of existing laws to AI scenarios, reflects an intent to protect society and individual rights without suffocating AI innovation and harming global competitiveness.

The policy is proactive with its guidelines that, while not legally binding, developers, providers and users of AI would be well-advised to follow. If an AI-related legal dispute arises, whether the parties followed guidelines may affect the outcome. While this “wait and see” strategy may seem reactive as AI is deployed in Japanese society, it is also flexible and can respond to real-world realities of AI in Japan.