— Reducing the Risks of US-China Dependence and Building a Transparent, Autonomous, and Competitive AI Foundation —
Original Japanese text is here.
The Open Source Group Japan (OSG-JP) is a non-profit organization dedicated to the promotion and development of open source within Japan. For a quarter-century, we have worked to contribute to the realization of an open digital society.
From the knowledge gained in this process, we are convinced that an Open Source approach is indispensable for ensuring transparency, collaboration, and sustainable innovation in the field of AI. Based on this conviction, OSG-JP was deeply involved in the joint design process of the “Open Source AI Definition” (OSAID) formulated by the Open Source Initiative (OSI), the international authority responsible for defining and approving Open Source licenses, and we have expressed our formal support for its final version. As we mark the first anniversary of OSAID’s official release, we believe that OSAID, as a global standard for an AI system to be truly recognized as “open source,” ensures the transparency and accessibility required by the evolution of AI technology and its ecosystem.
This proposal is based on the recognition that Japan’s national strategy for the development and dissemination of AI stands at a critical juncture in an era where AI technology fundamentally sways a nation’s industrial competitiveness and security. From our position and based on our expertise, we question the very approach that forms the core of technological development, in order to maximize the benefits of AI’s evolution while properly managing its risks. Furthermore, this proposal aligns at a high level with the “Principles of a Human-Centric AI Society” (deciaion by the Integrated Innovation Strategy Promotion Council) established by the Cabinet Office. We believe that positioning this Open Source international standard, OSAID, as the foundation of Japan’s AI policy will contribute to overcoming the challenges our nation faces and establishing leadership in the international community.
The Strategic Value of Open Source in the AI Era
Open source has transcended the framework of being merely a software license or a development culture that enjoys its benefits; it is now a proven economic model that forms the bedrock of modern digital infrastructure. According to joint research by Harvard Business School and the Linux Foundation, the “demand-side” value created by companies and individuals building products and services upon the value generated by Open Source communities is estimated to reach an astonishing 8.8 trillion US dollars.
Behind this phenomenal value creation lies the mechanism of the Open Source community: a dramatic reduction in development costs, the promotion of competition by lowering barriers to entry, and a collaborative framework of developers that transcends geographical constraints. This has generated innovation at a scale and speed that no single company or nation could achieve alone. This success model is expected to bring similar effects to the domain of AI development. Particularly in the development of foundational models, which requires massive capital, it prevents the market distortions and innovation stagnation caused by the technological monopoly of specific giant corporations. By fostering an open environment, it encourages the emergence of diverse models, leading to a healthy competitive landscape and continuous quality improvement. This, in turn, allows Japan’s startups, small and medium-sized enterprises (SMEs), universities, and public research institutions to gain fair access to cutting-edge foundational technology, directly contributing to the enhancement of Japan’s overall AI development capabilities and the diversification of its industrial structure.
Furthermore, as AI technology permeates every corner of society and begins to make judgments in critical areas such as medicine, finance, justice, and public administration, the transparency and accountability of its decision-making processes will become not only a technical requirement but also a critical social demand. At that time, “black box” AI, which humans cannot understand or verify, will find it difficult to gain social acceptance. Open source AI offers the most effective solution to this challenge. It enables independent third parties—such as public institutions, academic bodies, and civil society—to audit the internal workings of AI systems, identify and verify potential biases, and promote their correction. It brings a level of transparency and trust to society that proprietary AI can never achieve.
Thus, the promotion of open source AI is not merely a matter of technological choice. It is about building the very technical and social foundation for the “human-centric, responsible, and ethical AI society” that Japan aims to create. “Trustworthy AI” and “Open Source AI” are not merely complementary, but interdependent concepts.
Global Trends Surrounding Open Source AI
The movement to recognize the strategic importance of open source AI and place it at the core of national strategy is already accelerating in major countries around the world. The United States, the European Union (EU), and China, while taking different approaches, are aligned in their promotion of open source AI. For Japan to be left behind by this global trend would invite a situation that cannot be overlooked from the perspective of international competitiveness and security.
The United States laid out a comprehensive national strategy in “America’s AI Action Plan,” announced under the Trump administration, to win the global competition in the AI field. Within the “Acceleration of AI Innovation,” the first of the plan’s three pillars, “Encourage Open Source and Open-Weight AI” is clearly positioned as a policy item. Specifically, it recommends promoting the adoption of Open Source models by SMEs and expanding resources to make it easier for academic institutions and startups to access the computing resources, software, and data necessary for cutting-edge R&D. This approach clearly shows that the US views open source AI not as an object to be managed through regulation, but as a strategic asset to maximize market-driven innovation and enhance industrial competitiveness.
The European Union (EU), in its “EU AI Act”—the world’s first comprehensive AI legislation—adopts a “risk-based approach” that varies regulatory intensity according to the risks posed by AI systems. The handling of open source AI within this regulatory system is extremely informative. The regulation establishes exemption provisions that free AI systems released under Open Source licenses from many of the rules, as long as they do not fall into the category of AI with high, socially unacceptable risks. This strategic consideration avoids imposing excessive compliance burdens on Open Source AI development communities, including individuals and non-profits, and conversely, promotes innovation through open source AI. This “special treatment” is clear evidence that the EU recognizes the vital role of open source AI in balancing the dual goals of protecting citizen rights and safety through strict regulation and promoting innovation. In effect, the EU is using the regulatory framework itself to encourage the development of open source AI.
The People’s Republic of China has also publicly acknowledged the importance of open source AI at the national strategy level. In his keynote speech at the World Artificial Intelligence Conference in Shanghai in July 2025, Premier Li Qiang warned of the risks of “technological monopoly” in the AI sector and officially declared the “active promotion of Open Source AI development.” This statement, made shortly after the announcement of the US AI Action Plan, can be analyzed as a strategic move to position the concept of open source as an axis of opposition to technological dominance by specific countries or companies. Furthermore, China is advocating for “Chinese solutions” to support AI applications in developing countries and the establishment of international AI cooperation bodies, clearly signaling its intention to leverage open source as an important tool for expanding its international influence and shaping new standards in AI governance. It is an undeniable fact that under such national policies, Chinese-made AI models, distributed under licenses that are virtually Open Source, hold a significant share of the world’s open-weight AI models.
While the strategic intentions of these three parties differ—the US promoting innovation to win the AI market competition, the EU balancing innovation with strict regulation, and China countering Western technological hegemony and expanding international influence—the fact that these major powers are strategically converging on the promotion of open source AI based on their respective national interests holds extremely significant weight for Japan. In the global strategic competition over AI, for Japan to remain a mere passive observer or a consumer of Open Source AI ecosystems developed outside Japan risks inviting economic and technological subordination, wedged between this tri-polar structure of US market-driven, EU risk-based, and Chinese state-led approaches. We believe it is an urgent task for Japan to independently formulate and execute its own open source AI strategy—one that aligns with its national interests, industrial structure, and social values—and to secure a technology stack that can be verified, reproduced, and maintained domestically.
Proposals for Japan’s National AI Strategy
Based on these global trends, we propose that Japan strongly promote the following three policies as an integrated whole, clearly positioning them within its national strategy, in order to secure international competitiveness and build an AI society trusted by its citizens in the AI era.
Proposal 1: Promote OSI-Compliant Open Sourcing for AI Model Development, Including Foundational Models
A policy should be introduced that makes it a principle for AI development projects—particularly those for foundational models that could become social infrastructure—which are funded, supported, or procured by public institutions, to release their deliverables (source code, model weights and parameters, architectural information, training data, and related documentation) under a license compliant with the “Open Source AI Definition” (OSAID) formulated by the OSI.
Technological assets created with public funds should not be enclosed by specific companies or organizations but should be widely shared and utilized by society as a whole, becoming the foundation for further innovation. Thoroughly implementing this principle will maximize the return on investment of public funds and enable a wide range of domestic companies and R&D institutions, both large and small, to freely use, modify, and redistribute the results, thereby realizing the formation of a broad-based AI ecosystem. This is also an indispensable measure for avoiding dependence on specific giant vendors and maintaining Japan’s technological sovereignty. By adopting the internationally recognized OSAID standard, Japan’s efforts will facilitate collaboration with the global Open Source community and are expected to enhance international credibility. In addition, it is crucial to require a system for domestic redistribution, preservation, and continuous maintenance for all deliverables from publicly supported AI development, thereby ensuring operational continuity within Japan even if external services are suspended or their terms are changed. We are already seeing positive precedents for AI model development through industry-academia-government collaboration in Japan, such as LLM-jp led by the National Institute of Informatics, and these efforts should be further expanded.
Proposal 2: Implement Legal Protections for Developers to Sustain the Open Source Ecosystem
When formulating new policies and regulations concerning AI, Japan must explicitly avoid imposing regulations or requirements that would hold developers of Open Source AI components legally liable for outcomes resulting from the downstream use of the components they have developed and released. It must be legally clarified that liability lies not with the person who simply developed a general-purpose tool, but with the entity that implements and operates that tool for a specific purpose. This principle must also be thoroughly enforced in public procurement contracts. Furthermore, the government should not have the authority to unilaterally revoke the validity of Open Source licenses or force modifications.
The historical development of open source has been supported by the legal principles of “no warranty” (or “as-is”) and “limitation of liability,” which are typically included in licenses. Developers often release their work to the world as a good-faith social or academic contribution. If they were to be held responsible for damages resulting from the unforeseen use of their published work, they would retreat from development and publication, fearing the immeasurable legal risks. The same could happen to the Open Source AI ecosystem, and such a collapse must be avoided.
The importance of this legal principle is recognized even by the EU, which has a deeper understanding of the risks posed by AI, and it has intentionally reduced the burden on Open Source developers in the “EU AI Act.” Therefore, such legal protections are not mere preferential treatment for developers; they are essential infrastructure necessary to enable sustainable AI innovation in Japan. Correctly setting the focus of AI regulation—away from the general-purpose “tool creator” and onto the “tool user” who judges the risks and seeks to profit in a specific context—is the key to achieving both a healthy ecosystem and responsible AI use.
Proposal 3: Promote the Opening of Training Data to Foster Diverse AI Development
A national project should be strongly promoted to prepare and widely release high-public-interest data held by the government and related public agencies as open data, formatted as training datasets suitable for AI learning, all while strictly complying with the Act on the Protection of Personal Information and other relevant laws and regulations, and ensuring robust security.
The performance, characteristics, and potential biases of an AI model are fundamentally dependent on the quality and quantity of the data used to train it. Currently, one of the bottlenecks in AI development in Japan is the lack of diverse, high-quality, Japanese-language open datasets. If this challenge is left unaddressed, Japan will be forced to rely on AI models trained on overseas datasets, risking the proliferation of AI in society that does not reflect Japan’s linguistic, cultural, and social contexts or values.
A public-sector-led initiative to release reliable, large-scale datasets, especially those from domains specific to the Japanese-speaking world, would be the most effective means of breaking this impasse. This is expected not only to dramatically vitalize domestic AI development but also to promote the development of diverse AI applications that contribute to solving Japan’s unique challenges. America’s “AI Action Plan” also considers linking federal government funding to the public release of non-proprietary datasets by researchers, showing a shared international recognition of the strategic importance of data openness. Public data is a shared asset of the people, and utilizing it for AI, the intellectual infrastructure of the next generation, is an important responsibility of the government.
Conclusion
The promotion of open source AI and open data is not merely a technical policy choice. It is a central strategic imperative for establishing Japan’s economic security in the digital domain, reviving its international competitiveness, and, above all, building a democratic and open AI society that is trusted by each and every citizen.
As analyzed in this proposal, the United States, the EU, and China, based on their respective national interests and philosophies, have positioned open source AI as an important tool in their national strategies and are accelerating its use. Amid this major global trend, for Japan to remain a bystander is to accept de facto retreat, which could leave a significant liability for future generations. For Japan to demonstrate its unique strengths in the AI era and establish a solid presence as one of the world’s major players, it must shed its passive stance of merely consuming overseas ecosystems and adopt an active will to lead in building and driving an open ecosystem itself.
We, the Open Source Group Japan, strongly urge the Japanese government and related public institutions to swiftly implement the policies outlined in this proposal—namely, “the promotion of OSI-defined AI systems,” “legal protection for developers,” and “the opening of training data”—as an indivisible, three-part strategy.
Open Source Group Japan
Shuji Sado
Masayuki Hatta
