Ukraine shifts AI policy to self-hosted models after U.S. restricts Anthropic access

3 Sources

Share

Ukraine will prioritize AI systems it can run on its own servers, a senior official announced, as wartime Kyiv seeks independence from remote systems that providers can restrict. The policy shift follows U.S. government orders to Anthropic to cut access to powerful models, with Ukraine's Chief AI Officer calling AI sovereignty "a necessity, not just a defensive talking point."

News article

Ukraine AI Policy Shifts to Self-Hosted Infrastructure

Ukraine will favor self-hosted AI models that operate on its own servers, marking a strategic shift in the nation's approach to artificial intelligence deployment. Roman Kyslyi, Chief AI Officer at Ukraine's Ministry of Digital Transformation, confirmed the policy direction on Tuesday, emphasizing that wartime Kyiv must maintain control over digital tools used across government services, businesses, and the military

1

. The approach limits reliance on remote systems that providers can restrict or switch off, directly affecting AI models operated without provider control from companies like Anthropic and OpenAI whose main models remain under centralized control

2

.

AI Sovereignty Becomes Operational Necessity

The Ukraine AI policy was reinforced after the U.S. government ordered Anthropic to cut access to powerful models, a move that echoed broader European concerns about digital infrastructure dependency

3

. "It confirms that AI sovereignty isn't just a defensive talking point, it's a necessity," Kyslyi told Reuters. This statement reflects growing global tensions around AI access, with Reuters also reporting that Chinese authorities are considering curbs on top AI models that currently dominate the open-source AI models market

1

. The decisive criterion for Ukraine is not geographic origin but operational control. "If the vendor will provide it to run on our on-premise infrastructure, there are no restrictions," Kyslyi explained, adding that "the model is essentially a commodity" that Ukraine would work with from any provider whose technology could be deployed under Ukrainian control

2

.

Current Systems and Transition Plans

Ukraine's AI assistant inside the Diia government app currently runs on Google's remote-only Gemini model, accessed through servers within the European Union. Google provided free tokens for the service, meaning no budget spending for Ukraine

1

. However, Ukraine strips personal data before sending queries to the Gemini model because "they don't control those models," with Kyslyi describing it as an "interim" solution

3

. This temporary arrangement highlights the tension between immediate operational needs and long-term strategic goals around AI for government and military use.

Developing Indigenous AI Capabilities with Google's Open-Source Gemma

Ukraine is developing its own model in partnership with Kyivstar, based on Google's open-source Gemma, the open variant of Google's proprietary system. The on-premise model is due to be released in autumn and intended for use across government services, private enterprises, and the military

2

. Before selecting Gemma, the ministry compared several open-source options including Mistral models and OpenAI GPT-OSS. According to Kyslyi, both Gemma and Mistral models matched remote-only alternatives on many performance tests

1

. This evaluation process demonstrates Ukraine's commitment to balancing performance requirements with sovereignty concerns, ensuring that self-hosted AI models can deliver comparable capabilities to centralized alternatives while maintaining operational independence.

Implications for Global AI Deployment Strategies

The shift in Ukraine AI policy reflects broader concerns about U.S. government restrictions and the vulnerability of nations dependent on external AI infrastructure. For countries watching geopolitical AI dynamics, Ukraine's approach offers a template for maintaining digital autonomy without completely rejecting foreign technology. The emphasis on deployment flexibility rather than origin suggests that open-source AI models may become increasingly attractive to governments seeking to balance innovation with control. As more nations confront similar questions about AI sovereignty, the market for self-hosted solutions and open-source alternatives could expand significantly, potentially reshaping how AI companies structure their offerings for government and military clients worldwide.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved