NVIDIA and Palantir have announced a new sovereign AI capability built on NVIDIA’s open-source Nemotron models, enabling US government agencies and critical infrastructure operators to deploy, customise and continuously improve AI models within highly secure environments. The platform combines NVIDIA Nemotron open models with Palantir’s Sovereign AI Operating System, allowing organisations to retain full control over their data, model weights and deployment infrastructure. The system is designed for air-gapped and highly regulated environments where sensitive information cannot be connected to external networks. Agencies will be able to train AI models using their own operational data, retain ownership of the resulting models and continuously improve performance through internal feedback loops. The deployment is supported by NVIDIA AI Enterprise and Palantir’s Artificial Intelligence Platform (AIP), Foundry, Ontology and Apollo platforms. NVIDIA said the initiative reflects the growing importance of open AI models for government and enterprise development, arguing that they offer greater transparency, customisation and lower deployment costs than proprietary alternatives. The company also highlighted the role of open models in strengthening AI adoption across sectors including defence, healthcare, energy, transportation and public administration.