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

NVIDIA has detailed a group of European public-sector AI initiatives that rely on its open Nemotron models, spanning the European Commission, Poland, the UK, Germany, France and Greece. The projects share a common goal: letting governments run advanced models on infrastructure they control, adapt them to local languages, and keep sensitive data inside controlled environments. Several partners are also building country-specific editions of the Nemotron-Personas collection, the datasets behind NeMo User Sim, NVIDIA's open source library for simulating populations to evaluate and train models and agents.
At the European Commission, Berlin-based deepset is expanding its Haystack Enterprise Platform, which underpins AI@EC, the commission's shared platform for building and running AI applications on commission-managed infrastructure. The update links agent development, execution, optimization and governance for all customers and adds tooling to compare operating costs, assess answer quality and trace problems. NVIDIA Triton Inference Server supports the shared services used to find and rank relevant information. The expansion builds on requirements demonstrated by deployments such as AI@EC, where reusable components, role-based access, audit trails and guardrails support document summarization, knowledge management and multilingual policy analysis.
In Poland, Deloitte is helping build agentic systems for healthcare and employment services using open models. The eHealth Centre is developing an Intelligent Services Platform for medical imaging on NVIDIA DGX BasePOD with the MONAI framework, aimed at helping physicians analyze X-rays, CT scans and MRI images, detect abnormalities and prioritize urgent cases to cut waiting time for results. The Ministry of Family, Labour and Social Policy is developing an agentic factory on NVIDIA accelerated computing infrastructure, designed for more than 400 labor offices and running models including PLLuM, GPT OSS and Inkling for document search, summarization and digital assistants. NVIDIA is also working with SpeakLeash in Poland and the University of Ljubljana in Slovenia on country-specific additions to its synthetic personas program, with localized datasets planned to follow.
London-based Locai Labs released Nemotron-Personas-UK, a dataset of 1 million synthetic personas reflecting the UK's demographic and geographic diversity while reducing privacy risk. Suggested uses include healthcare evaluation and bias testing across age groups, education levels and rural and urban populations. For teams building public-service models, the appeal is a repeatable way to check whether a model behaves consistently across different communities without touching real citizens' data.
In Germany, Materna is introducing Sinfonie, an on-premises platform combining generative AI, voice interaction and 3D avatars, with NVIDIA Audio2Face available for animation. Built on a project for a German state administration, it lets employees query institutional knowledge by voice while data stays on the organization's own infrastructure. For classified work, secunet, NVIDIA, Red Hat and deepset are developing a sovereign cloud offering on secunet's SINA Cloud foundation, which includes security components authorized for use with classified information. The planned offering combines NVIDIA accelerated computing and Nemotron open models, Red Hat OpenShift AI for deploying and managing workloads, deepset's Haystack for agent orchestration, and infrastructure run by secunet in German data centers. The Sovereign Open Source Foundation Models initiative, known as Soofi, is meanwhile developing a family of models for German industry and the public sector, including Nemotron-Personas-Germany, an open dataset of 1 million records each carrying six synthetic personas, grounded in the 2022 census across all 16 federal states and 400 districts. It was created with NVIDIA NeMo Data Designer without real people's private data and can be used with NeMo User Sim, as can the UK edition.
In France, CANUT, the public purchasing body for digital services, selected LINAGORA and Docaposte to provide a sovereign collaborative workspace built on Twake.ai. More than 2,500 eligible organizations, including municipalities, schools and healthcare institutions, can use the framework without running their own tender. It offers email, document sharing and collaboration tools, with NVIDIA technology supporting speech transcription and infrastructure hosted in France; Région Occitanie is the preferred partner for scaling deployment. In Greece, KIEFER is developing an agentic factory on NVIDIA DGX SuperPOD with DGX GB200, alongside Sophea open weight models and Greek-language tools for knowledge retrieval, speech and digital assistants.
NVIDIA also ran an Open Models Codefest for startups, software providers and labs working on locally controlled public-sector solutions such as medical reasoning, AI tutoring and model governance. The top five teams were Pleias, Zehnlab.ai, Disseqt, Algoritmica GmbH and ABA Technology, with Deutsche Telekom and T-Systems providing compute for training and deployment.
The thread running through these projects is sovereignty as an architecture choice rather than a policy slogan: open weights, locally hosted inference, audit trails and synthetic evaluation data. In practice, a government that adopts Nemotron models can fine-tune them on local languages and inspect behavior before deployment, but it also takes on the operational burden of running and securing that stack itself. Synthetic persona datasets are a notable middle ground, since they let teams test for bias and coverage across regions and age groups without exposing residents' records; the trade-off is that a synthetic population is only as good as the census and design assumptions behind it. The German classified-cloud plan, which pairs SINA Cloud with OpenShift AI and Haystack, suggests vendors are betting that regulated buyers will pay for integration and certification rather than raw model capability. Timelines, pricing and general availability for the sovereign cloud are not laid out in the announcement.
Why it matters: European public-sector buyers are becoming a real market for open-weight models paired with on-premises infrastructure, which affects vendors selling orchestration, inference and evaluation tooling as much as model providers. Developers targeting government work should expect requirements around local hosting, auditability, data lineage and multilingual evaluation to be table stakes, and synthetic persona datasets such as the UK and German editions could become a standard way to prove a model works across a population. Inference: the emphasis on verifiable training data and governed deployment suggests procurement will increasingly reward transparency over benchmark scores alone.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Governments need modern solutions to serve the public while keeping sensitive information under their control. Open models provide a path forward, letting governments evaluate them, adapt them to local languages and public services, run them on infrastructure they control, apply their own safeguards and keep sensitive data within controlled environments. Some open models go further, […]