Japan's AI safety community is facing a critical funding crisis that threatens to dismantle years of progress in building expertise on existential AI risks. Despite having dozens of researchers, organizations, and conferences dedicated to AI safety, most groups operate entirely on volunteer labor with zero budget. This funding gap is causing talented researchers who have completed rigorous training programs to abandon the field and leave Japan, representing a significant loss of human capital and institutional knowledge.

Why Is Japan's AI Safety Funding Crisis Happening Now?

Over the past three years, Japan has built a surprisingly robust AI safety ecosystem. Since AI Safety Tokyo was established in January 2023, multiple organizations have emerged focused on education, governance, and technical research. The community has hosted roughly 80 AI safety events in the past year, run advanced training curricula, and even launched international conferences that have grown prestigious enough to relocate to Oxford.

Yet despite this visible progress, the infrastructure is crumbling from within. A detailed analysis by Esa Koskinen, a volunteer director at AI Safety Tokyo, estimates that Japan faces a funding gap of approximately $2.1 million USD and could employ AI safety researchers at roughly 1.8 to 2.3 times cheaper rates than in the United States. In 2027, Japanese organizations could theoretically hire around 35 researchers and 10 staff members with strong marginal effectiveness, but they lack the funding to make these hires happen.

What Organizations Are Struggling the Most?

The funding crisis spans the entire ecosystem. Here's the landscape of major groups and their current constraints:

• Shiba AI Lab: Employs roughly 35 researchers who have published at top machine learning conferences like ICML and ICLR, but operates with essentially zero budget and cannot convert these researchers to full-time positions.
• Independent Technical Researchers: About 25 researchers are conducting foundational and practical AI safety work, but lack funding and are dropping out of the field due to financial pressure.
• AI Safety Tokyo: Runs monthly study seminars and governance work with six board members but has no budget to expand operations or scale impact.
• AI Alignment Network: Operates six board members focused on awareness building for researchers through monthly panel talks, but cannot expand without funding.
• AGI Hub: Manages public awareness and existential risk discourse with only $1,250 per month to fund one employee, leaving the rest of the board unpaid.
• TARA Tokyo: Runs the ARENA curriculum in Japan with minimal funding, struggling with low public awareness and limited English-speaking student groups.

Even Japan AISI, the government-backed organization with a $56 million budget, faces a paradox. Despite its massive resources, the organization struggles to enable talented researchers outside its industry partnerships to access funding, leaving billions in potential impact on the table.

How to Address Japan's AI Safety Funding Bottlenecks

Koskinen identified four major bottlenecks preventing the field from scaling. Addressing these requires targeted intervention:

• Simplify Grant Applications: Reduce the time burden of applying for Western grants, which currently requires 50 to 200 hours of effort per application and discourages independent researchers from even trying.
• Improve Funding Visibility: Create centralized resources about available funding opportunities, since many researchers in Japan are unaware of grants they could pursue or lack familiarity with the application process.
• Increase Success Rates: Develop mechanisms to improve the odds of grant approval, as researchers currently face daunting prospects of submitting 10 or more applications before finding interest.
• Support Retention Programs: Create pathways for overseas researchers who come to Japan to conduct AI safety work to stay in the field, rather than leaving after six months when funding runs out.

Who Is Leaving and Why Does It Matter?

The retention crisis is particularly acute. Koskinen estimates that roughly a dozen people in his personal network have come to Japan from overseas, conducted independent AI safety research for six months, realized they could not secure funding, and left the field entirely. Extrapolating from conversations with roughly one-third of the local AI safety community, the real number is likely two to four times higher.

Many of these departing researchers have completed high-quality training programs like MATS (Machine Learning for Alignment Toolkit), ARENA (AI Research Experience for Alignment), SPAR, or TARA. Their departure represents a sunk cost in training and a wasted opportunity for the field. It is also deeply demotivating for the remaining community members who watch talented colleagues leave due to lack of resources.

Japan has significant appeal as a place to live and work. Koskinen noted that dozens of AI safety researchers have expressed a preference to live and work in Japan rather than the United States or London, if only the funding pathway were clearer and more realistic. This suggests that Japan could plausibly become Asia's largest AI safety hub with proper investment.

What Are the Broader Implications for AI Safety?

The geographic concentration of AI safety research in the United States and United Kingdom creates blind spots in how the field approaches existential risk. Japan brings unique perspectives shaped by its cultural emphasis on safety, caution with technology, and interest in machine consciousness. These viewpoints could help balance the techno-optimism that dominates Western AI discourse.

Additionally, Japan's political landscape presents an underexploited opportunity. Japanese politicians are under-educated on AI risks, especially existential ones, and rarely push back on issues regarding AI development. However, many decision-makers have expressed willingness to learn more. The shortage of Japanese-language materials on AI safety means that this receptive audience lacks the resources to develop informed policy positions.

The loss of Japan's emerging AI safety ecosystem would represent a significant setback for global AI governance. The 2026 relocation of the Technical AI Safety conference from Tokyo to Oxford, while a sign of the conference's prestige, also illustrates a failure mode: Tokyo did not have a funded ecosystem to retain and grow the conference it had created.

For funders and organizations committed to reducing existential AI risks, Japan represents a high-impact, neglected opportunity. The cost-effectiveness of funding AI safety research in Japan, combined with the region's strategic importance and cultural alignment with safety-conscious technology development, suggests that closing this $2.1 million funding gap could yield outsized returns for the field.