Japan's AI Act: A soft law approach to promoting AI development and utilisation

This article explores Japan's AI Act, recently passed by the National Diet on 28 May 2025, alongside other legislative measures aimed at fostering AI development in Japan.

Key points include:

• Unlike the EU AI Act, Japan's AI Act primarily outlines governmental responsibilities. It requires business operators to co-operate with governmental measures, such as guidelines and surveys, but does not impose legal sanctions for non-compliance. The government has indicated that it may issue administrative guidance or public warnings concerning malicious AI businesses. According to Diet discussions, this obligation extends to foreign entities.

• Japan's approach to AI regulation focuses on a soft law strategy based on guidelines designed to promote the development and use of AI. 

• Other laws in Japan are favourable to AI developers. Japan's Copyright Act offers flexible copyright exceptions that allow text-and-data mining without granting opt-out rights to right holders, even for commercial uses.

• Additionally, the Japanese government plans amendments to the Act on Protection of Personal Information aimed at facilitating the collection of learning data for AI development.

Japan's AI Act

On 28 May 2025, Japan's National Diet passed the AI Act, formally known as the Act on the Promotion of Research, Development, and Utilization of Artificial Intelligence-Related Technologies.

Unlike the EU AI Act which imposes strict regulations on businesses, Japan's legislation provides fundamental governmental policies for fostering AI while addressing associated risks.

The act outlines several governmental responsibilities including:

• establishing an AI Strategy Headquarters comprising all cabinet ministers to create basic policies for promoting R&D and utilisation of AI

• supporting research and development

• developing the infrastructure and datasets

• preparing guidelines

• securing and training individuals with AI expertise

• promoting education in AI

• conducting surveys and research, then implementing necessary measures based on the findings for research institutions, business operators, and the general public.

The sole obligation applicable to businesses under this act is to co-operate with policy measures from national and local governments. This includes following guidelines and partaking in governmental surveys and research. Notably, there are no legal penalties for failing to comply. However, during discussions at the Diet, the government indicated that it would issue administrative guidance to businesses engaging in malicious use of AI or publicly warn about specific types or names of harmful AI services.

While the extraterritorial application is not explicitly stated within the act itself, an Interim Report that forms its basis suggests that 'overseas business operators should also be covered by the legal system same as domestic business operators' (p7). During Diet discussions, the Minister in Charge of Science and Technology clarified that the obligation to comply with governmental measures under this act applies to foreign business operators.

Japan's approach to AI

Japan has opted for a soft law approach in governing AI by relying more on guidelines rather than stringent regulations. This strategy is based on the idea that agile governance through continuous improvements and amendments effectively manages rapidly evolving AI technologies.

Although Japan has introduced its own AI Act as mentioned earlier, it maintains a soft law stance without imposing stringent obligations on businesses. The core guidelines under this approach are encapsulated in "AI Guidelines for Business," jointly published by MIC (Ministry of Internal Affairs and Communications) and METI (Ministry of Economy, Trade and Industry). These guidelines outline measures developers, providers, and users of AI should undertake.

One driving force behind Japan's emphasis on promoting rather than regulating AI stems from concerns about its technological lag compared to other nations. In 2023, private investment in Japanese AI was merely USD 680 million—a stark contrast against USD 67 billion in the US, USD 8 billion in China, and USD 4 billion in the UK. Surveys also indicate a significantly lower adoption rate of generative AI among Japanese individuals and businesses.

The Minister in Charge of Science and Technology expressed ambitions regarding the enactment of Japan's AI Act, stating: “Our goal is to make Japan the world’s most AI-friendly country in terms of development and utilisation.”

Nonetheless, illegal uses of AI will still be penalised under existing laws, such as the Penal Code and sector-specific legislation. There have been cases where individuals using generative AI illegally were found guilty or arrested—for activities like creating ransomware or distributing obscene images.

Flexible copyright exceptions

Other aspects of Japanese law favour businesses involved in developing AI. The Copyright Act introduced 'flexible copyright exceptions' in 2018 which allow the exploitation of copyrighted works for purposes such as AI development without needing permission from copyright holders unless the enjoyment of copyrighted works is intended.

This provision resembles text-and-data mining exceptions within the EU’s Digital Single Market Copyright Directive but is more accommodating towards developers in terms that it does not grant opt-out rights to copyright holders even for commercial uses.

In March 2024, the Agency for Cultural Affairs issued a publication titled "General Understanding on AI and Copyright in Japan." This document addresses key topics related to the application of the Copyright Act concerning AI, including criteria for applying flexible copyright exceptions, issues surrounding copyright infringement from AI-generated outputs, and considerations about the copyrightability of AI-generated works.

Planned amendments to the Act on Protection of Personal Information

The government is currently planning amendments to the Act on Protection of Personal Information (APPI). The proposed changes include measures aiming to facilitate the collection of learning data for AI development. Currently, under the APPI, businesses must obtain consent from data subjects when acquiring sensitive personal information (eg, medical history) or providing personal data to third parties. The government's plan seeks to exempt this requirement in instances where personal information is statistically used for AI development. This may include scenarios where AI processes publicly available sensitive personal information as learning data or multiple business operators share and statistically analyse personal data to develop AI.

Initially, it was expected that the government would submit the amendment bill during the current Diet session. However, this submission has been postponed to allow for further negotiations among stakeholders. The business community has expressed concerns about introducing an administrative monetary penalty system and a class action system within the bill. It is anticipated that after addressing these concerns through continued discussions, the government will proceed with submitting these amendments.

This article was authored by Koshi Okabe.