Japan’s Strategy for Building a Robust Domestic AI Ecosystem

Artificial intelligence (AI) reach is widespread, affecting many sectors across public and private spheres, both positively and negatively. Through the Hiroshima AI Process, G7 members have recognized the real-life implications of both AI and generative AI’s immense potential.^[1] It can drive innovation, upskill workers, stimulate entrepreneurship, and improve healthcare outcomes and system innovation. However, the technology also introduces risks to human rights, privacy, safety and security, and can lead to disinformation and copyright infringements.^[2]

As the field of AI, particularly generative AI, continues to expand, countries worldwide grapple with two key challenges: regulating this rapidly evolving technology and fostering domestic AI development to reap its benefits. A robust domestic AI development sector is critical to maximize gains and implement these technologies for the greatest benefit to citizens. Harnessing the potential of this technology goes beyond technological prowess; it is linked to a nation’s future competitiveness, security, and societal well-being.

Japan, a key driver behind the G7’s Hiroshima AI Process and a leader in automotive and consumer electronics, understands this.^[3] Over the past two years, the country has taken a proactive approach, not only by contributing to global AI governance frameworks but also by creating a robust domestic AI ecosystem. This ecosystem focuses on integrating AI responsibly into society while fostering a solid environment for domestic development and adaptation of this new technology within its society and industry. In this article, I examine Japan’s approach to AI governance and the key characteristics of its AI ecosystem, particularly its emphasis on extensive stakeholder involvement and collaboration across government, industry, and academia to drive domestic AI development and innovation.

Japan’s Approach to AI Governance

Japan has consistently favoured the integration and utilization of AI, conceptualizing its integration into society through social principles.^[4] This means that regulations and guidelines surrounding AI in Japan have historically focused not on restricting AI, but on utilizing it for a society that promotes human dignity, diversity, and inclusion. This philosophy is outlined in the landmark document “Social Principles of Human-Centric AI” (2019), which details the country’s goals for utilizing AI responsibly.^[5] Currently, Japan does not have regulations restricting AI use. Instead, it has published a series of strategy and goal-based guidelines. Notable documents include “The AI White Paper,” Japan’s National Strategy for AI (2023),^[6] and the “Draft AI Guidelines for Business” (2024)^.[7] A key theme across these documents is the emphasis on principles and voluntary compliance by industry and citizens rather than a “hard law” approach with legally binding instruments and restrictions.^[8]

This “soft law” approach distinguishes Japan from countries like China, which rely on rule-based regulations with detailed standards, and requirements for AI systems.^[9] Japan prioritizes non-binding principles and recommendations over comprehensive regulations.^[10] This approach aims to avoid regulations becoming obsolete due to the rapid evolution of AI. Additionally, it fosters innovation by allowing flexibility to adapt to this changing landscape. Instead of a one-size-fits-all model, Japan employs an outcome-based approach. This means focusing on achieving measurable AI-related outcomes without defining specific or rigid compliance processes. This flexibility allows different entities to develop and utilize AI in multiple ways.

This unique regulatory landscape sets the foundation for Japan’s goals and its pursuit of becoming a global leader in AI technologies. Building on this framework is the country’s approach to AI innovation and adaptation, which involves extensive deliberation with multiple stakeholders, moving away from solely government-led initiatives.

Government as Facilitator

Recognizing the complexity of AI, the Japanese government positions itself as a facilitator rather than the sole creator of innovation. It acknowledges the private sector as the primary driver and mobilizes various government institutions and departments to support this. While the government steers away from sole-government initiatives, it mobilizes various government institutions and departments.^[11] One such department is the Ministry of Foreign Affairs (MOFA), which coordinated actions in the Hiroshima AI Process.^[12] Another example is the Cultural Affairs Agency, which has focused on examining AI and copyright issues, specifically data used in training sets and plagiarism concerns.^[13] Amongst government departments, a key department in Japan’s AI system is the Ministry of Economy, Trade, and Industry (METI), which has played a crucial role in developing Japan’s AI strategy documents and approaches. Beyond individual departments, task forces such as the AI Strategy Council^[14] (formed in 2023), and the AI Safety Institute,^[15] announced in January 2024, have also been established.

Funding and Investment

Funding is crucial for developing and stimulating the growth of AI and investment in domestic industries. In February 2024, METI launched the Generative Artificial Intelligence Accelerator Challenge (GENIAC).^[16] To finance GENIAC, METI secured ¥8.4 billion JPY with the goal that this investment will enhance domestic GenAI capabilities.^[17] This aid includes subsidies for data centres offering cloud computing and the development of supercomputers needed to process data for training AI models. Globally, start-ups and smaller developers struggle to enter the market due to limited access to infrastructure, such as computer capacity and cloud infrastructure. GENIAC, overseen by METI, bridges this gap through partnerships with domestic companies, start-ups, and international companies like Google LLC.^[18]

The Japanese government’s funding commitments extend beyond software research and development. It has also invested in chip production and digital infrastructure, both critical components of the AI supply chain. In November 2023, the government announced plans to secure ¥2 trillion JPY to support domestic chip production.^[19] Additionally, METI committed ¥45.5 billion JPY to data centre subsidies to establish centres outside Tokyo and Osaka from 2022 to 2026.^[20] Such investments allow for significant opportunities for Japan’s chip manufacturing industry and data centres. Moreover, investments not only offer the promise of profit but are also essential to boost the domestic AI start-up landscape, which positions Japan to establish Tokyo as an alternative to Silicon Valley.^[21]

Collaboration for AI Development

Research and development networks that bring together industry and academia are also a crucial element of Japan’s domestic AI ecosystem. An example of this is the AI Japan R&D Network (AI Japan), which has members from universities, research institutes, and private companies.^[22] It bridges stakeholders by recommending policies and facilitating research and development collaborations. Building on the AI Japan network is GenAI, a non-profit launched in 2024. This organization connects Japanese researchers, government officials, and industry partners like Google Cloud, Microsoft, and Oracle to support adoption, domestic development, technology sharing, and rule formulation for GenAI. Academia and industry collaborations are also paving the way for Japan’s homegrown AI development efforts.^[23] Examples include the Microsoft Base at Ritsumeikan University^[24] and Keio University’s collaboration with IBM on the Artificial Intelligence Utilization Project.^[25]

One characteristic of Japan’s adoption of AI has been the impetus to develop independent Japanese GenAI systems. Big Tech companies such as Microsoft, Amazon, and Google dominate AI supply chains, raising concerns about dependency, power concentration, and security. Building domestic AI capabilities offers two main advantages: it reduces the risk of data leaks and bolsters national competitiveness. Furthermore, Generative AI models like Gemini and ChatGPT are trained on English datasets, which have been criticized for not being able to capture the nuances of the Japanese language and context.^[26] Therefore, for Japan to have a robust domestic AI system to maximize benefits for its citizens, adaptation and the use of homegrown models are critical elements. Leading this effort are Japanese companies such as NTT, which has announced its LLM called “tsuzumi” and will offer it to businesses to improve efficiency and digital transformation starting in March 2024.^[27]

Challenges and Considerations

Despite ongoing initiatives, Japan still faces many challenges in its AI journey. One challenge is compliance and ambiguity. While the aforementioned soft law regulatory approach is geared towards supporting innovation, it relies on stakeholder agreement and buy-in. Another challenge is the country’s slow pace of digital transformation. Despite advancements in robotics, Japan ranks 32nd globally in digital competitiveness.^[28] Therefore, although the technology might exist, its broad implementation and uptake depend on the infrastructure and digital literacy of citizens. Japan also ranks 10th globally when it comes to newly funded AI companies. In 2022, the country only had 32 newly founded companies in this field, whereas in comparison, the US had 542.^[29]

Workforce disruption is another concern. According to the International Monetary Fund, 40% of global employment is exposed to AI, to which Japan is not immune.^[30] While the country utilizes AI as “Workforce 2.0” to address labour shortages amid its ageing population, the government will face challenges in balancing job security and AI deployment.^[31] Since the release of ChatGPT and other LLMs, there has also been a rapid increase in cybercrime. In Japan, the number of cybercrimes doubled in the first half of 2023 compared to the total for 2022. With this, the elderly population is most susceptible to fraud and phishing scams.^[32] Another concern has been brought forward by Japanese artists expressing concerns over copyright infringement.^[33] Japan amended its copyright act in 2018 to allow data scraping for the training of large models.^[34] The latest on this matter is that despite concerns, copyright law will not be enforced on material used to train models.^[35]

Lessons Learned and the Road Ahead

Japan’s AI landscape presents a dynamic yet complex picture. By examining its experience, we can glean valuable insights for navigating the intricate challenges of AI governance. Key takeaways from Japan’s AI ecosystem highlight the importance of a flexible regulatory approach, multi-stakeholder engagement, emphasis on R&D, and the development of domestic models. Overall, the field of artificial intelligence is constantly in flux. Balancing the needs of diverse stakeholders and navigating the tension between innovation and regulation requires adaptability. Moving forward, Japan’s success hinges on ongoing monitoring and evaluation to determine if its soft regulatory approach proves effective in achieving its AI goals.