Japan's Diet has passed the AI Innovation Framework Act, creating what industry analysts are calling the most permissive AI regulatory environment among major economies. The law explicitly permits AI training on copyrighted material, establishes a regulatory sandbox for AI companies, and introduces voluntary — not mandatory — safety standards.
The legislation passed with broad bipartisan support and takes effect in July 2026.
Training Data: No Restrictions
The most significant provision codifies Japan's existing copyright exception for AI training into permanent law. Under the new framework, AI companies can train models on copyrighted text, images, audio, and video without obtaining licenses or paying royalties, as long as the training is for the purpose of developing AI capabilities rather than reproducing specific works.
This directly contradicts the approach taken by the EU, which requires AI companies to disclose training data sources and respect opt-out requests from copyright holders. It also contrasts with ongoing litigation in the United States, where the legal status of AI training on copyrighted material remains unresolved.
Japan's position is that AI training constitutes a transformative use that benefits society broadly, and that restricting training data would disadvantage Japanese AI companies without meaningfully protecting creators.
Regulatory Sandbox
The law creates a three-year regulatory sandbox for AI companies that establish research and development operations in Japan. Companies within the sandbox receive:
- Expedited regulatory approval for AI applications in healthcare, finance, and transportation
- Tax incentives including a 30% R&D tax credit for AI development expenses
- Reduced compliance requirements during the sandbox period, with a transition to permanent standards after three years
- Government-funded compute access through a national AI infrastructure program
To qualify, companies must employ at least 50 people in Japan and commit to a minimum R&D investment threshold.
Industry Response
The response from major AI companies has been immediate. OpenAI announced plans to open a research lab in Tokyo — its first in Asia outside of a sales office. Anthropic confirmed it will expand its existing Tokyo presence into a full R&D center. Mistral AI said it is evaluating locations in Tokyo and Osaka for its first Asian office.
SoftBank, already Japan's largest AI investor, committed $15 billion to build AI data centers in Tokyo and Osaka to support the anticipated demand. CEO Masayoshi Son called the law "the single most important policy decision for AI in the last five years."
International Criticism
The law has drawn criticism from European regulators and content creators. The European Commission issued a statement noting that Japan's approach "undermines international efforts to establish fair compensation for creators whose work trains AI systems."
Hollywood studios and major music labels have also objected, arguing that Japan is creating a haven for AI companies to train on copyrighted content without consequence. The Recording Industry Association of America said it is exploring trade complaint options.
The Strategic Calculation
Japan's bet is straightforward: by offering the most permissive regulatory environment, it can attract the talent, investment, and infrastructure that will make it a global AI hub. The country has watched the EU's regulatory approach push AI development toward the US, and it is positioning itself as the alternative for companies that want fewer restrictions than either jurisdiction offers.
Whether this strategy succeeds depends on whether regulatory permissiveness alone is enough to overcome Japan's existing challenges in attracting international tech talent — including language barriers, work culture differences, and cost of living in Tokyo. The next 12 months will provide the first real data points.
