Ukraine to Select AI Models Operated Without Provider Control for Government Use

Ukraine will select AI models that operate without provider control to power government applications, a Ukrainian official told The Straits Times. The policy choice reflects Ukraine’s emphasis on AI sovereignty in the context of ongoing conflict.

The official made clear that AI models used by the government cannot be fully controlled by their providers. This means Ukraine is leaning toward open-weight models or special deployment arrangements that ensure model runtime is not subject to external interference.

This decision aligns with Ukraine’s broader strategy of digital infrastructure autonomy. Since 2022, Ukraine has been strengthening control over its digital infrastructure, and AI model selection is a natural extension of that strategy.

From a technical standpoint, models without provider control are typically open-weight models that can be deployed on user-owned infrastructure, giving users full control over runtime, fine-tuning, and inference. This contrasts with API-accessed closed models, where providers can modify, restrict, or terminate service at any time.

Ukraine’s decision could set a precedent for other nations. As geopolitical tensions rise, more countries may prioritize controllable AI deployment, potentially reshaping AI business models and global distribution patterns.

It remains unclear which specific models Ukraine will adopt, but leading open-weight options like Meta’s Llama series, China’s Qwen series, or DeepSeek series would be reasonable candidates given their cost and controllability.

This development highlights a critical emerging issue in global AI governance: control over AI models is becoming an integral component of national sovereignty and security.