Japan's AI Promotion Act — A Law Without Penalties and an AI Strategy Headquarters

Japan's AI Promotion Act—formally the Act on Promotion of Research and Development, and Deployment of Artificial Intelligence-related Technology (Act No. 53 of 2025, in the government's provisional translation)—is the first Japanese law to address AI directly. It was passed by the House of Councillors and enacted on 28 May 2025, and promulgated, with some provisions taking effect, on 4 June. It came fully into force on 1 September, including the provisions on the AI Basic Plan and the AI Strategy Headquarters. The Cabinet Office calls it simply the "AI Act".

Its defining feature is that it is a promotion law without penalties. Its purpose is to advance research, development and use; risks are handled by the state drawing up guidelines, investigating cases of harm and giving guidance and advice to businesses. That is a fundamentally different construction from the EU's AI Act, which fines violations.

How it became law

According to the House of Councillors' bill record:

Committees in both houses attached supplementary resolutions to the bill.

Structure

The law has four chapters and 28 articles, plus supplementary provisions.

Article 1 describes AI-related technology as a foundation for the development of Japan's economy and society, with the aim of improving people's lives and the sound development of the economy. The basic principles in Article 3 add that AI matters for national security as well as for efficiency and new industries, and stress keeping domestic R&D capability and raising the international competitiveness of the industry. Article 3(4) then states that, if used for improper purposes or in inappropriate ways, AI may facilitate crime, leaks of personal information, copyright infringement and other harms to people's rights and interests, and that measures such as ensuring transparency must be taken.

Handling risk: guidelines, investigation, guidance

Two articles carry most of the law's response to risk.

• Article 13 (ensuring appropriateness): the state prepares guidelines consistent with the intent of international norms. On this basis the AI Strategy Headquarters adopted the AI Guidelines on 19 December 2025. The relevant international norms include the code of conduct of the G7's Hiroshima AI Process, which Japan led.
• Article 16 (research and investigation): the state analyses cases in which people's rights and interests were infringed through improper or inappropriate AI development or use, and on that basis gives guidance, advice and information to research institutions and businesses.

Businesses that use AI are required to cooperate with national and local measures, while research institutions and citizens are only asked to endeavour to cooperate—a deliberate difference in wording. There is still no penalty for failing to cooperate.

The Headquarters and the Basic Plan

The AI Strategy Headquarters sits in the Cabinet. The Prime Minister is its head, the Chief Cabinet Secretary and the minister for AI strategy are deputies, and every other minister is a member—a design meant to run AI policy across government rather than ministry by ministry.

The first AI Basic Plan, adopted on 23 December 2025, set the goal of rebooting Japan through trustworthy AI and making Japan "the most AI-friendly country in the world" (in the government's provisional English translation). It sets out three principles—promoting innovation while mitigating risks, agile methods, and integrated domestic and international policy—and four basic policies: Adopt AI, Create AI, Enhance AI Trustworthiness and Collaborate with AI. On safety, it commits to roughly doubling the staff of the government's AI Safety Institute (AISI) immediately, benchmarking against the UK's AI Security Institute of about 200 people. A second plan was adopted on 14 July 2026.

Compared with the EU AI Act

Support and criticism

In the Diet, critics said the law did too little. According to the Japanese Communist Party newspaper Akahata, JCP lawmaker Tetsuya Shiokawa argued in the House of Representatives Cabinet Committee in April 2025 that 77% of respondents in a survey said AI needed regulation, and called for risk-based rules and stronger protection of rights.

The absence of penalties remained an issue after enactment. According to an April 2026 report carried by Asia News Network, the ruling Liberal Democratic Party's AI and Web3 subcommittee, chaired by former digital minister Masaaki Taira, drafted proposals urging the government to consider more effective measures, including penalties for unscrupulous operators that fail to respond to the government's requests for reports. The draft said damage from copyright infringement and deepfakes was becoming apparent, that effective means of addressing the challenges were limited under the current law, and that without penalties Japan might be unable to obtain information from AI operators that other countries can. It also pointed to foreign generative-AI services producing unauthorised images and videos closely resembling Japanese anime and manga characters.

The law itself anticipates revision: Article 2 of the supplementary provisions requires the government to review how it is working in light of international trends and to take measures if needed.

Social impact

The most direct change is inside government: Article 4(2) makes active use of AI in national administrative bodies a duty of the state. For companies, the law creates no new obligations or penalties, but it puts in writing their duty to cooperate and their exposure to guidance and advice when cases of harm arise.

For individuals, the law provides no direct remedy for someone harmed by AI. That is left to existing laws such as the personal-information and copyright acts, and to the state's investigation and guidance under Article 16. For workers, the Basic Plan lists "Collaborate with AI" as a basic policy, promising to transform industries and employment structures and develop skills, with details left to future revisions of the plan. Whether putting promotion ahead of penalties can both expand use and deter harm depends on how the guidelines are applied and on the review the law already foresees.

The wider story is on the AI timeline.