NVIDIA Open Models Power a New Wave of European Public Sector Projects

NVIDIA is leaning hard into a simple message: governments want modern AI, but they don’t want to hand over the keys to their data. Across Europe, it’s backing public-sector projects built around open models, local infrastructure and controls that stay inside national or organizational walls.

In Brussels, Berlin-based deepset is expanding Haystack Enterprise Platform for AI@EC, the European Commission’s shared system for building and running AI apps on commission-managed infrastructure. The new setup ties together agent development, execution, optimization and governance, and adds tools for comparing operating costs, checking answer quality and tracing problems. NVIDIA Triton Inference Server supports the shared retrieval services underneath it. The broader idea is familiar, but the mechanics matter: reusable components, role-based access, audit trails and guardrails are already part of deployments like AI@EC.

Poland is taking a more operational route. Deloitte is helping build agentic systems for healthcare and employment services, while the eHealth Centre is using NVIDIA DGX BasePOD and the MONAI framework for medical imaging. That system is meant to help physicians read X-rays, CT scans and MRI images, spot abnormalities and move urgent cases up the queue. Separately, Poland’s Ministry of Family, Labour and Social Policy is working on an agentic factory for more than 400 labor offices, using models including PLLuM, GPT OSS and Inkling for document search, summarization and digital assistants.

The pattern repeats elsewhere. In the U.K., Locai Labs released Nemotron-Personas-UK, a dataset of 1 million synthetic personas built to reflect demographic and geographic diversity while reducing privacy risk. In Germany, Materna is building Sinfonie, an on-premises platform with generative AI, voice interaction and 3D avatars, while secunet, NVIDIA, Red Hat and deepset are working on a sovereign cloud offering for classified information. France’s CANUT selected LINAGORA and Docaposte for a sovereign collaborative workspace, and Greece’s KIEFER is building an agentic factory with Greek-language tools. Different countries, same demand: useful AI, but on their terms.

That’s the real story here. Europe is not asking for bigger chatbots; it is asking for controllable ones. NVIDIA seems to understand that the procurement team, not the demo video, is the audience that matters.