Sakana AI Wins Award at US-Japan Competition for Defense Innovation

Sixty companies showed up for the US-Japan Global Innovation Challenge 2025, a competition run jointly by the Pentagon's Defense Innovation Unit and Japan's Acquisition, Technology & Logistics Agency. Only six made the finalist cut in each of the two tracks — biodefense and disinformation. Sakana AI, a Tokyo-based research lab, was the sole company to land a finalist spot in both, and walked away from the March 19 pitch event in Tokyo with the Innovative Spirit Award, one of just three handed out that day.

What's notable here isn't that Sakana AI beat specialists at their own game — it didn't win either category-specific award. It's that a company whose core business is AI agents and foundation models, not biosecurity or cybersecurity, managed to convince defense officials from two governments that its outsider perspective was worth recognizing anyway. Sakana AI says as much itself: most of the other entrants came in with deep domain expertise in biosecurity or cyber, while its own bet was that frontier AI research could offer something those specialists couldn't.

For the biodefense track, Sakana AI pitched a mix of tools built around a time-series foundation model for predicting infectious disease outbreaks, paired with an agent-based model that simulates human movement to forecast how an outbreak might spread, weigh countermeasures, and even help trace where it started. On the disinformation side, the company leaned on its existing work in image generation and diffusion models to build a detector for AI-generated images and video, one that runs fast and explains its own reasoning rather than just spitting out a verdict. It claims the system can flag outputs from Imagen, Midjourney, DALL-E, Flux, and Ideogram, and because it builds on existing foundation models rather than training from scratch, the company says it can keep pace as those generative tools keep improving.

Sakana AI frames the whole effort as an early step into defense work for a company that, by its own account, wasn't built for this sector. CEO David Ha told Japanese broadcaster TV Tokyo that Japan's defense industry is likely to weave AI into every layer of its technology stack over the coming years, which he sees as a long-term opportunity. Whether that optimism survives contact with the messier realities of biosecurity and information warfare is a separate question — but for one afternoon in Tokyo, judges from two defense agencies decided the bet was worth an award.