Ukraine is actively building a sovereign national artificial intelligence model to reduce reliance on foreign technology, strengthen national security, and modernize its public services.
Announced in late 2025, this initiative is part of a broader strategy to transition Ukraine from a “Digital State” to an “Agentic State,” where AI automates government processes and aids decision-making.
Key Development Partnerships
Ukraine has secured strategic partnerships with major U.S. technology firms to build this system, leveraging their existing tools while maintaining sovereign control over the final product.
- Google (Model Architecture): The national model is being built upon Google’s Gemma open-weight framework (specifically cited as Gemma 3 in recent reports). This allows Ukrainian developers to start with a powerful, pre-trained base rather than building from scratch, significantly accelerating development.
- Nvidia (Infrastructure & Hardware): Ukraine has partnered with Nvidia to establish the physical and technical infrastructure required for a “sovereign AI.” This collaboration gives Ukraine access to Nvidia’s advanced hardware and full AI software stack, which is critical for training and running the model securely.
- Kyivstar & Local Developers: The project is a joint effort involving the Ministry of Digital Transformation, the WINWIN AI Centre of Excellence, and Kyivstar (Ukraine’s largest telecom operator), ensuring the model addresses local business and consumer needs.
Strategic Goals and Applications
The model is designed to serve both civilian and military sectors, with a heavy emphasis on data security and independence from external contractors.
- The “Diia” Ecosystem: A primary application is the Diia AI LLM, a specialized version of the model tailored for Ukraine’s digital government platform, “Diia”. It functions as an AI assistant for citizens, helping them navigate legislation and access public services efficiently.
- Defense and Security: The military plans to integrate the model into battlefield systems for tasks such as drone coordination, enemy tracking, and data analysis. A locally controlled model ensures that sensitive military data does not leave Ukrainian jurisdiction.
- Cultural and Linguistic Sovereignty: Unlike global models trained primarily on English data, this sovereign LLM is specifically trained on Ukrainian data, including local dialects, legislation, and historical context. It supports multiple languages relevant to the region, including Crimean Tatar and Russian, to better analyze local information environments.
Infrastructure and Timeline
- Current Status: Initial training and operations are utilizing Google’s cloud infrastructure (Vertex AI) to speed up deployment.
- Future Plans: The roadmap includes a transition to local Ukrainian servers. This shift is intended to grant the Ukrainian government complete control over the AI systems used by its 23 million citizens, mitigating risks associated with foreign data hosting.