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

NVIDIA and its partners described a run of European public-sector AI projects built on open models and locally controlled infrastructure: the European Commission's shared AI platform, Polish healthcare imaging and labor-office assistants, German on-premises tools and a planned sovereign cloud for classified work, a French public collaboration workspace, and a Greek agentic system.

The common thread is control. Governments want to inspect models, adapt them to local languages, run them on their own hardware and keep sensitive data in-house. That's a reasonable demand, and it's reshaping how AI gets sold to the state.

We think sovereignty-by-design is the right default for public services. Citizens' health records and job-seeker files shouldn't be routed through a foreign vendor's API by accident. Open models that can be audited, and in some cases come with published training data so a government can check licensing and bias, are a genuine improvement over taking a vendor's word for it.

Some of these deployments carry real stakes, and that's where I want care. Poland's eHealth Centre is building a platform to help physicians read X-rays, CT scans and MRIs, flag abnormalities and prioritize urgent cases. Done well, that shortens waits. Done badly, triage software quietly deprioritizes the wrong patient. Poland's labor ministry is building an agentic system meant to support more than 400 labor offices. Assisting staff is fine; we'll be watching that it doesn't slide into automated decisions about people's benefits without clear appeal.

The synthetic persona datasets are interesting. Locai Labs released a U.K. set of 1 million synthetic personas, and the German Soofi initiative built one grounded in the 2022 census across all 16 federal states and 400 districts, made without real people's private data. Testing whether a public-service model works equally well for rural pensioners and city students is exactly what governments should do. The limit is that synthetic people are only as representative as the statistics behind them, and simulated users don't complain the way real ones do.

Then there's the word sovereign itself. Many of these systems are sovereign in where data sits and who operates the servers, while still depending on one American company's chips and software. That isn't hypocrisy, it's the current reality of the industry. But Europe should be clear-eyed that local hosting and strategic independence are not the same thing.

Named projects include deepset's Haystack platform underpinning AI@EC on Commission-managed infrastructure; a planned sovereign cloud from secunet, NVIDIA, Red Hat and deepset built on secunet's SINA Cloud, which includes components authorized for classified information; Materna's on-premises Sinfonie assistant in Germany; and a Twake.ai-based workspace selected by France's CANUT purchasing body that more than 2,500 eligible organizations can use without their own tender.