Japan's first binding AI statute is the Act on Promotion of Research and Development, and Utilization of Artificial Intelligence-related Technology, Act No. 53 of 2025. It was adopted on 4 June 2025, with the chapters establishing the AI Strategy Headquarters and the Basic Plan taking effect on 1 September 2025.
The statute contains no prohibited applications, no mandatory conformity assessment, no pre-launch registration and no monetary penalties on business operators. It creates governmental machinery and requests cooperation from industry. That is the whole design. Japan's stated position is that excessive regulation is itself a risk to be mitigated.
The operative document is the National AI Basic Plan, adopted by Cabinet on 23 December 2025 and revised in a second edition on 14 July 2026. It is built on four pillars: accelerating AI deployment, strengthening development capability, leading on AI governance, and building an AI powered society. The stated ambition is for Japan to become the most AI friendly country in the world.
Money follows the plan rather than the rulebook. The draft AI related budget for fiscal 2026 was reported at 502.7 billion yen, with roughly 90 percent of it, about 455.9 billion yen, directed at strengthening domestic AI development capability rather than at oversight. AI and semiconductors are designated top strategic fields under economic security policy, and Japan has set out a multi year industrial roadmap in which AI systems, semiconductor manufacturing and digital infrastructure absorb a large share of planned investment.
Soft law carries the behavioural load. The AI Guidelines for Business, updated to version 1.2 on 31 March 2026, consolidate expectations around human centricity, safety, fairness and transparency. They are not enforceable, and Japanese practice treats them as the reference point regulators will use when existing sectoral law is applied to an AI system.
That is the part outsiders tend to miss. A promotion statute with no penalties does not mean an AI system in Japan is unregulated. Product safety law, financial services law, medical device law, competition law and personal information law all still apply. What Japan declined to build is a separate horizontal compliance regime stacked on top of them.
The bet is legible: if the binding constraint on national capability is development speed rather than deployment harm, then the marginal yen is better spent on compute and research than on assessment bureaucracy. The European Union made the opposite bet. Japan is the cleanest live test of the first one.
For companies operating across both, the practical consequence is that the same model can ship in Tokyo under general law while the same deployment in Frankfurt triggers a conformity process under the EU AI Act. Governance teams that write one global policy tuned to the strictest regime will overpay in Japan. Teams that tune to Japan will fail in Europe.