Sakana AI Enters Defense Intelligence: Partnering with Japanese Think Tank DEEP DIVE to Advance AI-Powered Intelligence Analysis

Core Event: Sakana AI and DEEP DIVE Forge AI Intelligence Analysis Partnership

Japanese AI startup Sakana AI recently announced the signing of a formal partnership agreement with the general incorporated association DEEP DIVE. The two organizations will engage in deep collaboration in the field of AI-driven intelligence analysis. This partnership marks Sakana AI's official establishment of "defense and intelligence" as one of its core strategic directions.

What Is DEEP DIVE? A New Force in Japanese Civilian Intelligence Analysis

DEEP DIVE is a civilian intelligence organization co-founded by two prominent experts in Japanese military and international affairs — Bonji Ohara and Yu Koizumi. The institution possesses extensive professional knowledge in security and intelligence analysis, while also commanding diverse open-source data resources including satellite imagery.

Yu Koizumi currently serves as Associate Professor at the University of Tokyo's Research Center for Advanced Science and Technology, specializing in Russian security policy and military strategy research. After the outbreak of the Russia-Ukraine war in 2022, he became a public figure through his professional yet accessible military analysis on Japanese media, and has authored multiple books including The Geopolitics of "Imperial" Russia. Bonji Ohara formerly served as a Japan Maritime Self-Defense Force officer before transitioning to think tank research, conducting long-term studies on Chinese military modernization at institutions such as the Sasakawa Peace Foundation, with deep insights into Chinese naval strategy and Indo-Pacific regional security dynamics. The complementary expertise of the two — one focused on Russia, the other on China — precisely covers the two major geostrategic security concerns Japan currently faces, giving DEEP DIVE comprehensive regional analysis capabilities from its inception.

The DEEP DIVE organization they co-founded is positioned to fill the gap in Japan's civilian intelligence analysis capabilities. Notably, Japan has long been highly dependent on government agencies (such as the Cabinet Intelligence and Research Office and the Defense Intelligence Headquarters) and intelligence sharing from its American ally, with extremely weak independent civilian analysis capabilities. The emergence of DEEP DIVE represents Japan's formation of a dual-track government-civilian intelligence analysis ecosystem.

Partnership Objectives: Using AI to Break Through the Scale Bottleneck of Human Intelligence Analysis

Deep Integration of AI Technology and Professional Intelligence Knowledge

The core logic of this partnership is clear: combining DEEP DIVE's professional data and analytical experience with Sakana AI's proprietary AI technology to achieve analytical scale, speed, and precision previously unattainable through human effort alone.

Traditional intelligence analysis is highly dependent on human analysts' experience and judgment, but when facing massive volumes of open-source intelligence data (OSINT), human capabilities have inherent limitations in processing speed and coverage. OSINT (Open Source Intelligence) refers to the methodology of systematically collecting and analyzing intelligence from publicly accessible information sources, including commercial satellite imagery (such as high-resolution remote sensing data from Maxar and Planet Labs), public posts on social media platforms, flight and vessel tracking data (ADS-B, AIS signals), government public documents, academic papers, news reports, and more. During the Russia-Ukraine conflict, the OSINT community accurately predicted military operations multiple times before official confirmation by analyzing geographic markers in TikTok videos and changes in military deployments visible in commercial satellite imagery — fully demonstrating the enormous potential of open-source intelligence analysis while also exposing the efficiency bottleneck of human processing of such massive data volumes.

AI technology intervention promises breakthroughs in the following dimensions:

• Scale enhancement: Simultaneously processing multi-source heterogeneous data, including satellite imagery, social media, news reports, etc. For example, Planet Labs produces satellite imagery covering the entire global land area daily — full-volume analysis through human visual interpretation alone is simply impossible, while computer vision models can automatically detect changes in imagery and flag anomalies
• Speed leap: Compressing analysis cycles that previously required days down to hours or even minutes. In time-sensitive scenarios such as military conflicts, the timeliness of intelligence directly determines its value
• Precision enhancement: Discovering subtle patterns and correlations that human analysts might miss. The cross-lingual information integration capabilities of large language models enable simultaneous processing of intelligence sources in Russian, Chinese, English, and other languages, uncovering implicit correlations across multilingual information

Joint Research Roadmap: Iterative Human-Machine Collaboration

Both parties will advance the sophistication and practicality of analytical methods through continuous exchange of ideas. This means the collaboration is not simple technology outsourcing, but rather an iterative joint R&D process — AI models need continuous optimization through expert feedback, and experts also need to learn how to more effectively leverage AI tools.

This human-machine collaboration model is particularly critical in the intelligence analysis domain. Purely automated analysis is prone to false positives, and in the security domain, the cost of misjudgment can be extremely high. Therefore, the ideal model is for AI to handle large-scale data screening and pattern recognition while human experts handle final judgment and decision-making — the so-called "human-in-the-loop" architecture. DEEP DIVE's expert team is perfectly positioned to provide high-quality labeled data and feedback signals for the AI system, helping models understand the professional context of intelligence analysis.

Sakana AI's Strategic Landscape: From Fundamental Research to Defense and Finance Dual-Engine Drive

Sakana AI was founded in Tokyo in 2023 by former Google Brain researcher David Ha and Llion Jones (one of the co-authors of the Transformer paper Attention Is All You Need). The company name derives from the Japanese word for "fish" (魚), symbolizing its core technical philosophy — AI architecture inspired by the collective intelligence of fish schools in nature. Its technical approach emphasizes combining and optimizing multiple small models through evolutionary algorithms and swarm intelligence methods, rather than purely pursuing parameter scale expansion. This "nature-inspired computing" methodology gives it a unique position in Japan's AI ecosystem and potential advantages in resource efficiency.

Previously, Sakana AI was primarily known for its fundamental model research and academic papers, with its published research on model merging and evolutionary model optimization attracting widespread attention. The explicit designation of "defense and intelligence" alongside "finance" as priority areas now signals the company's acceleration from a research-oriented organization toward applied deployment.

This strategic choice carries significant implications. In the global AI competitive landscape, defense and intelligence is one of the application domains with the highest technical barriers and greatest added value. For a Japanese domestic AI company, entering this field involves both commercial considerations and a willingness to contribute to national security capability building. Japan's revised National Security Strategy at the end of 2022 explicitly called for strengthening intelligence collection and analysis capabilities, listing AI technology as a key enabling technology. The "Innovation Technology Research Institute" established under the Acquisition, Technology & Logistics Agency in 2023 increased investment in frontier technologies such as AI and quantum computing. Meanwhile, Japan's Ministry of Defense is advancing the "Multi-Domain Joint Operations" concept, in which rapid intelligence fusion and distribution is a core capability requirement. This policy environment provides institutional space for civilian companies like Sakana AI to participate in the defense sector, also reflecting the Japanese government's strategic intent to leverage civilian innovation to compensate for insufficient official technical capabilities.

Industry Trend: OSINT+AI Is Becoming the Global Standard for Intelligence Analysis

From a broader perspective, AI-empowered open-source intelligence analysis has become a global trend. America's Palantir Technologies was founded in 2003 by Peter Thiel and others, initially funded by In-Q-Tel, the CIA's venture capital arm. Its Gotham platform has long served the U.S. intelligence community and Department of Defense, helping analysts discover hidden correlation patterns by integrating massive heterogeneous data into visual analysis interfaces. Palantir's market capitalization exceeded $100 billion in 2024, reflecting capital markets' strong endorsement of the AI+defense track. Israel's Voyager Labs and Cobweave focus on social media intelligence analysis, while the UK's BAE Systems Applied Intelligence holds advantages in communications intelligence.

Japan started relatively late in this field, but as the geopolitical environment changes — particularly the rapid modernization of China's military forces, the normalization of North Korean missile threats, and the reshaping of the global security landscape brought by the Russia-Ukraine conflict — demand for intelligence analysis capabilities is surging. Japan previously relied primarily on intelligence sharing mechanisms with its American ally (such as peripheral cooperation with the "Five Eyes" alliance), with relatively lagging autonomous analysis capability development.

The Sakana AI and DEEP DIVE partnership can be viewed as an important signal of Japan catching up to international standards in the AI+intelligence analysis domain. Its effectiveness will depend on whether AI technology can truly understand and adapt to the professional context of intelligence analysis, and whether both parties can establish an efficient human-machine collaboration model. Particularly noteworthy is whether Sakana AI's evolutionary model combination approach can demonstrate unique advantages in the complex scenario of intelligence analysis, which requires multimodal understanding (text, images, geospatial data, etc.).

Summary: The New Organizational Form of AI+Intelligence Analysis Deserves Attention

The significance of this partnership lies not only in technological innovation exploration, but also in its representation of a new organizational form — deeply binding frontier AI capabilities with domain experts to jointly address complex real-world challenges. This model differs from traditional client-vendor outsourcing relationships, more closely resembling a joint laboratory or strategic alliance, with both parties forming a symbiotic relationship in intellectual property and research outcomes.

From the broader perspective of AI industry development, this also confirms an important trend: competition in general-purpose large models is giving way to competition in "AI + deep vertical domain integration." In high-barrier domains like intelligence analysis, pure technical capability alone is insufficient to form competitive advantage — the true moat lies in deep understanding of domain knowledge and accumulation of high-quality professional data.

For practitioners focused on AI application deployment and defense technology development, the Sakana AI and DEEP DIVE partnership is a benchmark case worth continuous tracking. Its success or failure will provide important reference for the universal question of "how AI startups can enter high-barrier vertical domains."