Estonia and Owkin Partner on Sovereign AI for Health Research

Owkin is partnering with Estonia’s National Applied Research Center, Metrosert, to explore how the company’s K Pro AI scientist can support biological and health research using sensitive national data inside secure Estonian infrastructure.

The collaboration centers on a sovereign-AI model: approved researchers would use advanced biomedical AI without requiring protected health information to leave the national environment established to govern it. The partners say the approach could enable reproducible analysis and faster generation of scientific insights while preserving Estonia’s control over its data.

The first phase will focus on selected Estonian health-data cohorts. Owkin and Metrosert will also examine research areas including oncology, where linked clinical records, molecular data, medical imaging and treatment outcomes could support future biomedical studies.

Bringing AI to nationally governed health data

Estonia has spent decades building a highly digitized public infrastructure and now holds population-scale health and biomedical datasets. The new project is intended to test how those foundations could become a reusable, AI-enabled research capability rather than a collection of data assets used only for individual studies.

Metrosert is a state-owned applied research organization whose work includes health data, autonomous vehicles, drone technologies, hydrogen technologies and biorefining. Its role gives the collaboration a national research partner responsible for connecting scientific development with Estonia’s broader industrial and public-sector priorities.

Liina Vahtras, Estonia’s Minister of Economic Affairs, said the country has already built the trust, digital infrastructure and health-data capabilities needed for responsible innovation. She described the next step as finding ways for those assets to produce greater value for science and future health innovation, while strengthening Estonia’s position in secure health-data use.

For Estonia, the potential benefits extend beyond a single research program. A secure national environment that can support advanced biomedical AI could help domestic researchers undertake more ambitious studies while making the country more attractive to international life-sciences organizations seeking access to well-governed, population-scale datasets.

What K Pro brings to the collaboration

K Pro is Owkin’s AI scientist platform for pharmaceutical research and strategic decision-making. It combines agentic AI, biomedical tools, multimodal patient data and scientific workflows in one research environment designed to operate with sensitive information.

Researchers can use natural-language questions to initiate work involving scientific literature, clinical information, pathology, genomics, transcriptomics and other data types. Owkin positions the platform as an assistant to scientific teams, helping organize complex analyses while researchers remain responsible for shaping hypotheses, reviewing evidence and deciding how findings should be used.

That distinction matters when AI operates against national health data. Large datasets are valuable only when their records are standardized, documented and governed well enough for researchers to understand how an analysis was produced and repeat it. The Estonia project will therefore test not only K Pro’s analytical capabilities, but also the data preparation, permissions and validation processes needed for national-scale biomedical AI.

“Estonia has built one of the world’s most advanced digital foundations,” said Thomas Clozel, CEO and co-founder of Owkin. He described the partnership as an effort to show what comes next: using sovereign AI to turn national biomedical data into scientific discovery, stronger research capability and ultimately better patient outcomes.

Oncology offers a demanding first use case

Oncology is a natural area for the partners to explore because cancer research often requires investigators to connect information stored across different systems. Clinical histories, pathology images, genomic profiles, molecular measurements and treatment outcomes can each provide part of the picture, but combining them consistently remains difficult.

An AI research environment capable of reasoning across those layers could help scientists identify patterns, generate hypotheses and prioritize follow-up work. Reproducibility is equally important: useful results must be traceable to the underlying data and analytical steps rather than delivered as opaque answers.

Owkin has been expanding K Pro through a series of pharmaceutical collaborations. In September, Boehringer Ingelheim licensed K Pro and multimodal patient data for discovery work in oncology and immunology. Owkin has also announced agreements involving AstraZeneca, Sanofi and Servier.

The Estonian collaboration expands that model from company-level research environments to national infrastructure. Instead of transferring a defined dataset into an external AI system, the partners will evaluate how the AI scientist can work where the data is already secured and governed.

A test of sovereign biomedical AI

The collaboration remains exploratory. The announcement does not identify the individual cohorts, provide a deployment timetable or define measurable research milestones. Those details will determine how quickly the project moves from infrastructure evaluation to active scientific programs.

Even at this stage, the project reflects a broader change in biomedical AI. The most important advances will depend not only on model capability, but on whether AI systems can function inside real institutional environments with appropriate permissions, provenance and validation.

Owkin and Metrosert are now testing that premise at national scale. If the model proves effective, Estonia could provide an early reference for governments considering how to make securely governed health data useful for AI-assisted research without weakening national control over the information.