Ukraine Wants AI It Can Control Itself, Official Says

Ukraine has decided to favor AI systems that it can run on its own servers, a top government official said on Tuesday, July 7. The country is at war with Russia and wants to make sure its digital tools keep working no matter what. If Ukraine depends on AI systems controlled by outside companies, those companies could cut off access at any time. That would put government services, businesses, and even the military at serious risk.

Roman Kyslyi is the Chief AI Officer at Ukraine's Ministry of Digital Transformation. He told Reuters that Ukraine will look for AI models that can be placed on Ukrainian computers and servers. This approach is called 'on-premise,' which means the software runs locally instead of on a distant company's machines. Ukraine wants to avoid systems that stay under the provider's control, such as the main AI models from companies like Anthropic and OpenAI.

Kyslyi said the policy became even stronger after the U.S. government told Anthropic to stop giving access to some of its most powerful AI tools. That move showed Ukraine why relying on outside providers is risky. 'It confirms that AI sovereignty isn't just a defensive talking point, it's a necessity,' Kyslyi said. AI sovereignty means a country has full control over the AI tools it uses.

Kyslyi made clear that Ukraine does not care where an AI model comes from, as long as it can be run on Ukrainian systems. He described AI models as a 'commodity,' meaning they are more like a basic resource than a special product. Ukraine will work with any company that allows its technology to run under Ukrainian control.

Right now, the AI assistant inside Ukraine's Diia government app uses Google's Gemini system. Gemini is a remote-only service, which means it runs on Google's faraway servers inside the European Union. Google has been providing this service for free, so Ukraine has not had to spend money on it. However, Kyslyi called Gemini an 'interim,' or temporary, solution because Ukraine does not fully control it.

Because Ukraine does not control Gemini, the government removes personal information from any questions before sending them to the system. This protects the private details of Ukrainian citizens. Kyslyi said this extra step is needed because they 'don't control those models.' Ukraine is working toward a safer, more permanent option.

Ukraine is now building its own AI model together with a company called Kyivstar. The new model will be based on Google's open-source tool called Gemma, which anyone is allowed to use and change. It is expected to be released in autumn and will be used by government offices, private businesses, and the military. Before choosing Gemma, Ukraine compared other open-source options, including models from Mistral and an open version of OpenAI's GPT, and tests showed Gemma performed just as well as many remote-only systems.