Sakana AI signs demonstration deal with Japan's defense ministry for intelligence analysis AI

AI moves into the work of intelligence analysts

Japanese AI startup Sakana AI has teamed up with the country's defense ministry. The company announced on August 24 that it signed a contract with the Ministry of Defense on July 29 covering "research and demonstration of AI functions needed for integrated analysis work." Under the deal, Sakana AI's AI agent technology will be applied to the work of intelligence analysts at the Defense Intelligence Headquarters.

The Japanese government has named "intelligence capability" as one pillar of its broader push to strengthen national defense. As the security environment grows more complex, the volume of information that needs to be handled has grown too — and officials say using AI to speed up and sharpen collection, analysis, and management has become an urgent priority.

What the demonstration covers

Sakana AI's work on this project will focus on three fronts: making intelligence collection more efficient, improving analytical capability, and building more systematic information management. All three build on the company's AI agent technology, and the ultimate goal, as the company explains on its announcement page, is to embed AI functions reliable enough to actually support real-world decision-making.

How this differs from March's contract

This isn't Sakana AI's first contract in Japan's defense sector. Back in March, the company won a research contract from the Acquisition, Technology and Logistics Agency's Defense Innovation Science and Technology Institute to develop foundational technology for upgrading command-and-control (C2) systems. That project focused on integrating information from units and sensors in the field to enable faster, more accurate command decisions.

This new contract is a different kind of work. Rather than command-and-control at the unit-operations level, it targets strategic intelligence work that underpins national security policy as a whole. From battlefield-level decisions to national policy-making, Sakana AI appears to be steadily expanding its footprint in strengthening Japan's intelligence capabilities.

Hiring expands into defense

Sakana AI is also growing its team to support this area. The company says it's hiring across multiple roles, including software engineers, applied research engineers, cybersecurity engineers, project managers, and technical editors. It also stated that strengthening national security infrastructure with domestic AI technology is an urgent priority.

Editor's view

What stands out about this deal is the sequence. Sakana AI entered the defense sector in March with unit-level command-and-control technology, and has now moved up to intelligence analysis work that supports national policy decisions. Startups expanding into defense work typically follow a path that starts with field-support technology and moves toward policy-decision support — Sakana AI made that jump in just five months. It's officially a demonstration project, but the fact that the company landed a second contract with an organization that functions like a national intelligence agency suggests a trust relationship has already been established.

The trend of private AI startups entering defense and intelligence work isn't unique to Japan. Big U.S. labs have signed contracts with the Department of Defense too, and the push to protect security infrastructure with domestic technology is happening in multiple countries at once. What sets this case apart, though, is that the contractor isn't a company selling a large language model outright — it's a relatively young company that has focused specifically on agent technology. Work like intelligence analysis, which involves layered judgment and procedure, tends to suit agent-based workflows better than pure generative AI in practice, so the outcome of this contract could become a reference point for defense authorities in other countries as well.

For Korean companies, there are two things worth watching here. One is that AI demonstration projects for government and public institutions aren't the exclusive territory of large cloud providers. The other is that these contracts aren't built around selling a finished product in one shot — they're built through incremental trust, step by step. If similar attempts emerge in Korea to apply AI agents to defense and security work, starting with a narrow demonstration and expanding scope from there, the way Sakana AI has done, is likely the more realistic approach.

In the coming months, Sakana AI stands a good chance of landing a third defense contract built on the results of this demonstration. If it delivers results in intelligence analysis, the next stage would move beyond demonstration into actual deployment.