Ukraine officially begins development of its own large language model (LLM), which will become the foundation for national AI services in the public and private sectors.
This project is the first step towards building Ukraine’s sovereign digital infrastructure in the era of artificial intelligence (AI). Its goal is to provide the state, businesses, and citizens with effective, secure, and Ukrainian-made AI solutions.
The project involves close coordination between the state, business, and the scientific community. Its implementation will be managed by technical, coordination, and ethics committees. In addition to developing LLM, there are plans to launch more than ten AI products, including an AI assistant in Diia and tools for analyzing regulatory and legal acts. More about Ukrainian artificial intelligence.
Why does Ukraine need its own AI?
On June 18, 2025, an event took place in Ukraine that could be decisive in shaping the country’s digital independence: the launch of a project to create a large language model (LLM) with sovereign status was announced. The project will be implemented jointly by the Ministry of Digital Transformation and Kyivstar.
“This day will be important for our country as a whole. As you know, about six months ago, we adopted a strategy for innovation in our country. This is the first group document that we have been developing for more than a month,”
said Minister of Digital Transformation Mykhailo Fedorov.
According to him, despite the full-scale war, Ukraine has no right to remain on the sidelines of the global technological race. The world does not stop: it invests in human capital, develops companies, and creates added value: “We remain in this race, and we need to fight the aggressor while developing the economy, our own weapons, education, and human capital.”
Fedorov noted successes in the areas of GovTech and DefenseTech. Ukraine has risen from 102nd to 5th place in the global ranking of digital public services, and the number of companies in the defense technology sector has grown:
“We have made a breakthrough in the DefenseTech sphere. We have gone from 0 companies to 500, of which at least 10 are innovative players.”
The minister emphasized that no innovation strategy can exist without AI, because AI will change our world just as the internet changed it in its time. Every company will use AI. Some business processes will disappear, and entire industries will be transformed.
Ukraine intends to become one of the world’s top three leaders in terms of practical AI implementation. To this end, the WINWIN competence center has already been created, an AI vision is being formed, and investments in infrastructure are being made. Separately, Fedorov spoke about the creation of a sovereign LLM model in cooperation with Kyivstar: “It will allow us to store data in the country, launch services without the risk of leakage, and work with unique Ukrainian data sets, such as Delta.”
The model will be trained exclusively on Ukrainian sources, without distortions or hostile IPSOs. Applications range from government services to business, from references in Diia to defense sector analytics. These are our libraries, our scientists, our history. “All information will be digitized and used to train the model,”
added the head of the Ministry of Digital Transformation.
The project involves the creation of strategic, technical, and ethical committees. Duration – about 9 months. Upon completion, the model will become open-source for government and educational institutions.
Kyivstar: the operator’s participation
Kyivstar CEO Alexander Komarov spoke during the presentation about the company’s participation:
“We are undergoing the biggest transformation in the company’s history: from a telecommunications operator to a digital service provider. Kyivstar already has successful projects in television, healthcare, and mobility. Participation in the creation of LLM is a logical step in the development of competencies.”
Komarov recalled that the international group VEON, which includes Kyivstar, already has experience in creating an LLM model in Kazakh in collaboration with the Barcelona Computing Center:
“It was an 80 billion token model. It now works in Kazakh, English, and Turkish. We want to implement a similar approach in Ukraine.”
The project in Ukraine will also contribute to national security and sovereignty: “I believe that any country, from a national security standpoint, should have its own LLM model. It means independence, lower costs, and new opportunities.”
At the time of the announcement, the beta test of the first AI product in Diya is already underway — a chatbot that provides advice on services. Other services will join it in the future. The Ministry of Digital Transformation’s competence center is already analyzing regulatory and legal acts using AI, freeing up thousands of hours of lawyers’ work.
Funding, security, and commercial participation: what is known about the creation of a sovereign LLM model in Ukraine
Following the high-profile announcement of the creation of a large language model (LLM) in Ukraine, as part of a partnership between the Ministry of Digital Transformation and Kyivstar, speakers answered key questions from journalists about costs, security, commercial prospects, and the openness of the project to other players.
According to Mykhailo Nestor, Director of Digital Product Development at Kyivstar, the project budget has not yet been finalized due to the research nature of the initiative:
“No one has ever done this before — an LLM in Ukrainian. We are in talks with cloud partners, we have our own cloud business, and this will be another branch of it.”
He clarified that this does not involve a large team, but rather “10–15 highly qualified specialists,” as well as additional teams for data processing, ethics, and testing: “As many as we need to make a great project. We will not stop halfway.”
According to him, there is no point in saving money for a mediocre product, just as there is no point in making a product that is worse than Open Source or paid analogues.
When asked about Kyivstar’s commercial interest, Nestor replied that the company is interested in developing its own expertise in AI:
“We want to be as much of a leader in AI as we are in Data Science. This costs more than selling the model. The model will have national status and will be trained on Ukrainian data sets. For the state, defense, or finance, these will be separate tracks of use.”
Oleksandr Bornyakov, Deputy Minister of Digital Transformation, said that during the testing period, the model will be free for government and educational institutions.
During the discussion of future scenarios for using the model, representatives of the Ministry of Digital Transformation confirmed that one of the options for its application will be a chatbot in the Diya app. According to Deputy Minister of Digital Transformation Oleksandr Bornyakov, there will be an AI system in Diya at that time. “Because the team is currently working on it. And we could just include it in this model to see how it works,” Bornyakov explained.
We are talking about including the future LLM model in the existing functionality, in particular as part of a jam (trial launch). The main goal is to gather feedback on the quality of the system:
“When the first version of the model is ready, it will be important for us to get feedback on its quality. It is possible that testing on a government product will be one of the ways to identify what needs to be improved before the model becomes a full-fledged open-source solution.”
Thus, the public chatbot available to citizens is part of the strategic plan to test and improve the LLM model before its full release.
Mykhailo Nestor confirmed that other companies can join the project: “We are open. We can invest together.”
Danylo Tsivok, Chief AI Officer at the Ministry of Digital Transformation, added that the process of creating the model will not be limited to nine months: “This is a long-term process that involves improving data, digitization, and attracting new partners.”
When asked about the safety and combating of “hallucinations” in LLM models, Danylo Tsivok explained:
– the infrastructure will be certified;
– personal data will not be included in the training sample;
– hard rails will be used – filters for incoming and outgoing requests;
– An ethics committee will be in place to monitor biases.
– Human Evaluation will be conducted to assess the quality of results.
“This is a top priority. Safety is important both in the development process and in use,” added Tsivok. Regarding the size of the Ukrainian corpus, Tsivok acknowledged that it is not even sufficient for a medium-sized model: “We will need to make an open call, an initiative to collect additional data. Private organizations can also join in.” At the same time, all data must be collected in compliance with copyright and legislation.
Nestor and Bornyakov also confirmed that the model will be integrated into government and business services: “It will be like the internet. It’s not a choice — it’s a wave. It will be in all products, companies, and government agencies.“ Tsvyok made the following prediction: ”If 22 million users make over 100 million requests every year, the economic effect will be enormous.”
artificial intelligence / ChatGPT / copyright / intellectual property
