By Leo Marchandon and Gianluca Lo Nostro
Dec 1 (Reuters) – Ukraine is developing a large language model using Google’s open-weight Gemma framework, aiming to create an independent system for artificial intelligence operations amid rising demand across military and civilian sectors, the country’s digital ministry and mobile operator Kyivstar said on Monday.
The project will use Google’s computing infrastructure for initial training before shifting entirely to local infrastructure, ensuring Ukraine retains full control over AI systems accessed by 23 million citizens daily.
“The Ukrainian LLM is expected to serve as the foundation for a new generation of AI-powered services across the public and private sectors in Ukraine,” Kyivstar said in a statement.
Ukraine wants to build its own AI model instead of paying millions of dollars to foreign companies, Deputy Minister for Digitalisation Oleksandr Bornyakov had told Reuters in September.
Bornyakov said avoiding reliance on proprietary systems, such as OpenAI’s ChatGPT, was deliberate, explaining that Ukraine’s military intended to integrate AI into battlefield management systems for troop coordination and enemy monitoring. Chinese LLMs, including DeepSeek and Qwen, were also rejected, he added.
Ukraine’s Minister for Digitalisation Mykhailo Fedorov said in August the military already employed AI for aerial and satellite reconnaissance and drone piloting. AI tools from U.S. firm Palantir Technologies are helping analyse Russian strikes, monitor disinformation campaigns and determine demining priorities.
Gemma was chosen for its performance across multiple languages, a document shared by the ministry showed.
“Google was selected after extensive evaluation, by reinforcing the technological and economic ties between Ukraine and the United States—ties that were further deepened by Kyivstar’s Nasdaq listing in August 2025,” Kyivstar said on Monday.
Meta’s Llama and France’s Mistral AI models were also among the options being considered, two sources familiar with the matter told Reuters.
The project stems partly from communication gaps affecting current AI systems. When Bornyakov speaks to people from his hometown of Bolhrad in Odesa Oblast, they use a blend of Ukrainian, Russian and Bulgarian — a dialect existing AI models struggle to handle, he said.
“As soon as you start real production applications, you find these limitations very fast,” said Misha Nestor, Kyivstar’s chief product officer overseeing the project. He highlighted flaws like mistranslations in legal documents and AI-generated “hallucinations”.
The consortium has established four advisory committees with binding authority over technical, legal, cultural, historical and linguistic aspects of the new model, ensuring it handles Ukrainian and minority languages such as Crimean Tatar, alongside Russian, used widely in Ukraine.
Data collection from over 90 government institutions is underway, including court registries, educational publishers, regional archives and records of Russian actions during the ongoing war.
Training will occur on secure graphics processing units outside Ukraine, provided by Google, before the finished models are deployed on local data centres. Kyivstar was not immediately available to confirm the targeted launch date.
The eventual launch will face security concerns as Russian cyberattacks continue. “We understand immediately after release it will be attacked,” Bornyakov said. Other AI services have also faced apparent attacks, he noted.
Measures against threats, such as “prompt injection” attacks — malicious instructions embedded in tasks provided to the AI — are being developed.
Kyivstar, which became the first Ukrainian company to list on a U.S. stock exchange in August, has installed more than 3,500 backup generators to stabilize operations as Russian strikes on Ukraine’s energy infrastructure intensify.
The completed AI system will first support government applications and Kyivstar’s platforms before expanding to the private sector. Officials say the project demonstrates how smaller nations can leverage open technologies to achieve strategic independence in AI, countering reliance on dominant foreign systems.
(Reporting by Leo Marchandon and Gianluca Lo Nostro in Gdansk, editing by Milla Nissi-Prussak)
Comments