What we learned about AI sovereignty in Lithuania

By Adam Dustagheer and Jake Holland, Datnexa

In April, we joined a Digital Leaders delegation to Lithuania to explore how governments are approaching AI, data, and digital transformation in a rapidly shifting geopolitical and technological landscape. What we saw was not a blueprint to transfer to the UK, but a set of practical observations about how sovereignty in AI is being interpreted and operationalised in a smaller, highly focused nation.

Sovereignty is more than infrastructure

A recurring theme throughout the visit was that AI sovereignty cannot be reduced to owning data centres or developing foundation models. Lithuania’s approach recognises that sovereignty depends equally on the capability to use these assets effectively.

This means investing across three dimensions: hardware, software, and - critically - skilled people. Without the talent and institutional capability to deploy, adapt, and govern AI systems, infrastructure alone offers limited strategic value. Sovereignty, in this sense, is as much about applied capability as it is about technical ownership.

National context enables speed

Lithuania’s scale and structure create conditions that enable faster progress. Compared to the UK, there are fewer legacy systems and fewer organisational “super tankers” to turn.

We observed:

• More homogenised public systems, including more centralised data governance.

• A smaller, more agile vendor ecosystem, with SMEs such as Insoft playing a significant role in delivery.

• Shorter decision-making cycles, allowing policy and implementation to move in closer alignment.

These factors combine to create an environment where experimentation and iteration are more achievable at pace.

Regional integration shapes strategy

Lithuania’s geopolitical context is impossible to ignore. Proximity to Russia and Belarus, alongside strong ties with Nordic and Baltic partners, has created what can best be described as a “war-footing” mindset.

This translates into:

• Active investment in sovereign infrastructure, including national server capacity.

• Strong regional cooperation, particularly in security and data resilience.

• Engagement with EU-wide initiatives, including access to shared sovereign AI capabilities such as foundation models.

Sovereignty here is not isolationist. It is layered - national capability reinforced by regional alignment.

Data strategy as a foundation

Lithuania’s progress is underpinned by a deliberate approach to data. Consolidation and standardisation are treated as prerequisites for effective AI.

Examples include:

• The development of a National Data Library to centralise access and governance.

• The use of interoperable standards, controlled vocabularies, and shared ontologies.

• The creation of engineered datasets, similar to our National Frailty Index, combining local context with structured data to generate actionable insight.

This focus moves beyond data availability to data usability.

Life beyond hyperscalers

While hyperscale cloud providers remain part of the ecosystem, Lithuania is actively building value in the layers above them.

A key principle is portability. Sovereignty is strengthened by the ability to move workloads between providers such as AWS and Google, reducing dependency on any single supplier. This approach balances the benefits of hyperscale infrastructure with the strategic need for flexibility and control.

Reusable patterns, not rigid borders

One of the more compelling ideas we encountered was the notion that sovereignty does not require rigid national boundaries in technology design.

Reusable patterns, whether in data models, governance approaches, or AI deployment frameworks, can be shared across countries. This increases resilience while still allowing nations to retain control over their specific implementations.

The value of knowledge exchange

Finally, the delegation itself reinforced the importance of international collaboration. Across conversations with public sector leaders, technologists, and SMEs, there was clear common ground on both the opportunities and challenges of AI.

These exchanges do more than share ideas. They build practical networks that enable faster learning, reduce duplication of effort, and accelerate adoption of proven approaches.

Observations, not prescriptions

It is important to stress that Lithuania’s model is shaped by its size, history, and geopolitical context. The UK operates at a different scale, with different constraints.

However, the underlying principles - capability alongside infrastructure, data standardisation, ecosystem diversity, and strategic portability - offer useful lenses through which to consider our own approach to AI sovereignty.

The question is not how to replicate Lithuania, but how to interpret these lessons in a way that fits the UK’s unique environment.