Sakana AI proposed new AI solutions in both the biodefense and disinformation countermeasure categories at a competition jointly hosted by the US Department of Defense’s Defense Innovation Unit (DIU) and the Japanese Ministry of Defense’s Acquisition, Technology & Logistics Agency (ATLA). Out of 60 companies from around the world who participated, Sakana AI won one of the three awards at the competition.
The US-Japan Global Innovation Challenge 2025 is a competition jointly hosted by the Japanese Ministry of Defense’s Acquisition, Technology & Logistics Agency (ATLA) and the US Department of Defense’s Defense Innovation Unit (DIU). The goal of the competition is to promote defense innovation by identifying companies capable of developing novel technologies that can help advance biodefense solutions and counter disinformation.
Participating companies had a month to develop and submit their proposals for one of two themes related to biodefense or countering disinformation. Experts from the US Department of Defense and the Japanese Ministry of Defense reviewed the proposals submitted by 60 companies, and selected 6 finalists for each of the two themes. On March 19, 2025, the finalists were invited to the final pitch event in Tokyo.
Final Pitch Event for US-Japan Global Innovation Challenge held in Tokyo on March 19, 2025. (TV Tokyo, WBS)
Sakana AI was the only participating company to be selected as a finalist in both themes, and also received one of three awards at the competition, the Innovative Spirit Award. We were also the only Japan-based company to win an award at the event.
David Ha, CEO of Sakana AI, receives the Innovation Award from Kyosuke Matsumoto, Director of Technical Strategy at the Japanese Ministry of Defense’s Acquisition, Technology & Logistics Agency (ATLA). (TV Tokyo, WBS)
As a new entrant into the defense sector, Sakana AI has devised new AI solutions for both themes: detecting and diagnosing biological threats, and determining the accuracy of information and combating disinformation.
We developed and submitted our solutions to both biodefense and disinformation themes of the challenge, and we were the only company selected to be a finalist in both categories. While our proposals were not selected to win one of the two category-specific awards, we were confident that our solutions provide a unique, novel perspective to complement traditional approaches, and in the end, we took home the Innovative Spirit Award.
Official announcement from ATLA about the awards, and the companies selected in the finalist round for both challenges. Sakana AI is the only company selected to be a finalist in both categories.
Sakana AI’s Proposed Solutions
Sakana AI is a Japanese AI research lab, backed by NVIDIA, and several Japanese institutions. Our goal is to develop AI solutions for Japan’s needs. While many of the participants in this challenge were companies with deep expertise in biosecurity or cybersecurity, our core focus is in AI research, and pushing the frontiers of what AI agents can do.
For example, last year Sakana AI developed The AI Scientist, an AI agent technology that automates the scientific discovery process, and received international acclaim. We believe that through our research into AI agents and efficient foundation models, we can provide new, unique perspectives and contribute to innovation in fields such as biodefense and countering disinformation.
Theme 1: Detecting and Diagnosing Biological Threats (Biodefense)
In our solution for the biodefense theme, we proposed a comprehensive set of solutions for the detection and diagnosis of biological threats, using frontier AI models and AI agents.
In addition to using our time-series-based foundation model to predict infectious disease outbreaks, we also proposed solutions that combines an agent-based model that predicts human behavior with generative AI to accurately predict infectious disease outbreaks, identify options for countermeasures to curb the spread of infection and their impact assessment, and help identify the origin of an outbreak.
Demonstration of our Agent Based Model of population movement for predicting infection spread in Tokyo.
Sakana AI has extensive expertise in AI image generation and diffusion models, and we have developed our own systems for detecting AI-generated images. For this challenge, we have developed a state-of-the-art model that detects images and videos generated by diffusion models with high accuracy and low inference speed, and also proposed a solution that displays the explanation for the model’s decision.
Our solution can detect images generated by the latest AI-generated models such as Google’s Imagen, Midjourney, OpenAI’s DALL-E, Flux, and Ideogram. In addition, since our approach leverages existing underlying foundation models rather than being trained from scratch, we can efficiently improve our model to keep up with the latest in image and video generative models as they improve over time.
We tested our approach with real photographs or generated images related to wars, propaganda, disasters, etc. that are being spread on social media. For example, our model can identify AI-generated images like the ones below:
Demonstration of our AI-Image Detection model developed for the Disinformation Challenge. Our system provides the user with both the likelihood and explainability of whether an image is AI-generated. (TV Tokyo, WBS. March 19, 2024)
Accelerating Defense Innovation for Japan
Japan, being one of the most advanced and culturally significant countries in the world, is facing many crucial challenges this century. These challenges include facing a declining population, decreased competitiveness and increasing geopolitical tensions. As the leading democracy in Asia, Japan also faces the challenge of strengthening its defense infrastructure and capability.
While AI alone will certainly not solve all of Japan’s problems, we believe AI can help.
Sakana AI aims to produce AI innovations that can help Japan cope with challenges that it will face for decades to come. We believe that this challenge is the first step for Sakana AI to help accelerate defense innovation in Japan.
“Japan’s defense industry is expected to incorporate AI into all levels of the technology stack, leading to opportunities over the long term,” says Sakana AI’s CEO David Ha, who was interviewed by TV Tokyo’s WBS programme, where he discussed his views regarding sovereign AI, geopolitics, government and defense trends for the next decades.
We believe that this challenge is the first step for Sakana AI to help accelerate defense innovation in Japan.
Sakana AI
Interested in joining us? Please see our career opportunities for more information.