Ukraine is building its digital independence—one AI model at a time. In a landmark move blending national pride and cutting-edge innovation, Kyivstar and Ukraine’s Ministry of Digital Transformation have chosen Google’s Gemma as the foundation for the country’s first-ever national large language model (LLM). But here’s where it gets even more interesting: this project isn’t just about technology—it’s about sovereignty, security, and identity in the digital era.
Dubai and Kyiv, December 1, 2025 — VEON Ltd. (Nasdaq: VEON) announced that Kyivstar (Nasdaq: KYIV; KYIVW), in partnership with the WINWIN AI Center of Excellence under the Ministry of Digital Transformation of Ukraine, has officially selected Google Gemma as the base model for developing Ukraine’s own national LLM. The project will harness the computing infrastructure of Google’s Vertex AI to drive training and deployment at scale.
A new chapter in Ukraine’s digital autonomy
Kyivstar will take the operational lead in developing the Ukrainian LLM. The goal? To build an advanced language model that fully captures the richness of Ukrainian dialects, terminology, and cultural context—while keeping all sensitive national data securely stored and processed within the country. This is vital for government, finance, and healthcare systems where data protection is paramount.
Choosing Google as a partner followed months of evaluation and strategic discussion. The decision symbolizes a deepening relationship between Ukraine and the United States—strengthened even further when Kyivstar went public on the Nasdaq in August 2025. Opting for an open-source foundation from a global tech leader allows Ukraine to combine international expertise with national priorities, ensuring its LLM reflects both local authenticity and global standards for AI safety and transparency.
Voices from leadership
“We’re building our national LLM on a proven open-source platform,” explained Danylo Tsvok, Chief AI Officer at the Ministry of Digital Transformation and CEO of the WINWIN AI Center. “The next step is to train it on uniquely Ukrainian data. Our focus was on how well the model already understands Ukrainian-language materials and how easily we can control its further training. This approach limits linguistic inconsistencies and reduces ethical risks.”
VEON Group’s CEO, Kaan Terzioglu, called the move a transformative milestone: “Kyivstar and the Ministry are advancing Ukraine’s digital capabilities in a historic way. With a sovereign Ukrainian LLM, local users—from citizens to businesses to public institutions—will finally access AI tools that truly speak Ukrainian and understand Ukraine. Our mission is to deliver augmented intelligence built not just on text, but on genuine local context.”
What happens next
Once complete, the Ukrainian LLM will serve as the technological backbone for a wide range of AI-driven applications spanning both public and private sectors. Examples include legal and regulatory automation, personalized learning systems, medical research tools, and financial analytics platforms. By training the model exclusively on Ukrainian data, developers aim for results that are more accurate, relevant, and practically useful than generalized global models.
Krzysztof Kaziów, Director of Customer Engineering CEE at Google Cloud, praised the collaboration: “We’re honored that Kyivstar and the Ministry have chosen Gemma for this landmark project. Gemma’s multilingual strengths and balanced performance-to-resource ratio make it ideal for national-scale LLMs. We’re proud to support a project that enhances Ukraine’s digital landscape.”
Kyivstar’s technical roadmap begins with refining Gemma to better capture nuances of the Ukrainian language, optimizing its tokenizer, and training it on curated datasets. Dedicated benchmarks will continuously assess its accuracy, ethics, and adaptability for sector-specific applications.
Bridging AI divides across nations
This initiative builds on VEON’s ongoing strategy to close the AI language gap across its global markets. Following Kazakhstan’s KazLLM and Pakistan’s Urdu-language model, the Ukrainian LLM is VEON’s next bold step in creating locally grounded AI ecosystems. Each project strengthens regional digital sovereignty and gives communities more meaningful access to trustworthy AI.
About Kyivstar
Kyivstar Group Ltd., traded on Nasdaq, operates JSC Kyivstar—Ukraine’s leading digital operator and the first Ukrainian company listed on a U.S. stock exchange. The company offers a wide portfolio of services across mobile connectivity, broadband, cloud computing, cybersecurity, e-health, digital TV, ride-hailing, and enterprise data solutions. Together with VEON, Kyivstar plans to invest USD 1 billion in Ukraine between 2023 and 2027, targeting infrastructure growth, technological upgrades, and strategic acquisitions. Learn more at https://investors.kyivstar.ua. Nasdaq tickers: KYIV; KYIVW.
About VEON
VEON is a global digital operator delivering integrated connectivity and digital services to nearly 150 million users and over 120 million digital subscribers. Active in five countries—home to more than 6% of the global population—VEON uses technology to drive connection, growth, and empowerment. VEON is listed on NASDAQ. For details, visit https://www.veon.com.
Forward-looking statement
This release contains forward-looking statements as defined under the U.S. Securities laws (Section 27A of the Securities Act of 1933 and Section 21E of the Exchange Act of 1934). These statements involve potential risks and uncertainties that could cause actual outcomes to differ materially, including the scope, timing, and results of the Ukrainian LLM’s development. Investors should review the “Risk Factors” section in VEON’s 2024 Form 20-F filed with the SEC on April 25, 2025, and other filings for further details. VEON assumes no obligation to update forward-looking statements unless required by law.
Contact:
VEON – Hande Asik, Chief Communications and Strategy Officer
Email: pr@veon.com
And here’s a thought to leave you with: Should nations prioritize developing their own language models to protect cultural and data sovereignty—or does global collaboration in AI lead to faster progress? What do you think—where should the balance lie?
