Milestone Systems has entered a partnership with DCAI that will see its Vision Language Model hosted on the Gefion supercomputing platform, giving European developers a sovereign environment for building video-driven AI applications.
The move aligns closely with ongoing work around the AI Act and data governance, providing a route to smart-city deployments that remain inside the EU regulatory framework. In fact, Sovereign AI video analytics refers to video-driven artificial intelligence that is developed, trained, deployed, and managed entirely within the legal and infrastructure boundaries of a specific region or country. In Europe, this means AI systems that operate under EU jurisdiction, use data governed by GDPR, and run on European compute resources rather than relying on infrastructure outside the region.
The model itself has been developed with NVIDIA, using Cosmos Reason to analyse video and convert it into written output – summaries, event descriptions, and real-time alerts. For transport agencies and urban authorities, this could reduce the burden of manual review across vast camera networks. Milestone sees particular value in city-wide traffic management, airport operations, port logistics, and other settings where continuous visibility is essential.
With Gefion now in the mix, the VLM can be trained and deployed within a European cloud environment. “Our mission at DCAI is to provide the best AI infrastructure that does not compromise on data sovereignty,” said Dr Nadia Carlsten, CEO, “and we are proud to see companies building on top of that foundation to advance responsible AI initiatives.”
Milestone comes to the table with its Hafnia Data Library, built to support compliant training and transparent data handling. “This collaboration is a powerful step toward ensuring that Europe’s AI future is built on trust, transparency, and technological sovereignty,” said Thomas Jensen, CEO of Milestone Systems, “and by combining Hafnia, DCAI’s secure cloud, and NVIDIA’s AI capabilities, we’re helping create a uniquely European ecosystem for high-impact AI development.”
For Andrew Burnett, Interim CTO at Milestone, the technical uplift is a key part of the story. “The fine-tuned Vision Language Model represents a leap forward in video understanding and contextual AI for smart-city use,” he said, “and running it on an EU-based cloud keeps data, training, and deployment under European control.”
The intention is to give municipalities, researchers, and industry a scalable platform for building AI-enabled public infrastructure. With increasing focus on sovereign compute and responsible AI, this partnership adds a tangible step toward Europe’s stated ambition to lead in regulated, ethically-aligned AI development.
Benchmark will track how this progresses, particularly as real-world deployments emerge. If you’re working with sovereign video AI or planning a pilot, we’d welcome your insights for future coverage.