DXC Technology delivered a sovereign, strictly secured generative AI platform to the European Space Agency that keeps sensitive data exclusively within the ESA environment while also accelerating research and collaboration. The solution, built on Mistral models and industrial AI orchestration, shows that artificial intelligence can be deployed responsibly, manageably, and cost-effectively.
Why ESA needed sovereign AI
ESA processes vast amounts of sensitive information—from technical documentation of rockets and satellites to patents, security protocols, and strategic decisions. It wanted to use generative AI to accelerate development, innovation, and knowledge sharing, but the data could not leave the agency’s strictly regulated environment. The requirement, therefore, was sovereign AI with a high level of security, regulatory compliance, and rigorous cost control.
DXC was selected as the integrator and implementation partner. The company approaches AI as “customer zero”—it first develops and tests its solutions internally, fine-tunes processes and data, and only then deploys them to clients, saving time and budgets and minimizing the risks of experimentation for customers.
What the solution looks like
The Security/Sovereign AI platform was developed over more than two years in several waves and now supports thousands of experts. It is built on European Mistral language models and comprises three pillars: RAG as a service, the specialized chatbot Aska, and an enterprise “chatGPT” environment, all in ESA’s private operations. Collaboration with technology partners (such as Nvidia) and containerization approaches enabled scaling and fine-grained control of access rights.
The core of the solution lies in data, processes, people, and governance, not in chasing the latest model. AI orchestration at industrial scale delivers repeatable and auditable procedures, while the customer retains full control over data, infrastructure, and operations.
Results and next steps
More than 4 000 rocket specialists, data scientists, and other top professionals use the solution, and together with 11 000 other employees they constitute the main target groups. ESA described the project as highly successful; an approximately 30 % increase in productivity was recorded in development, administration, decision-making, and knowledge sharing.
DXC delivered consulting, data classification, integration, and AI application development in an “in time, in budget” mode and under strict security conditions. The platform is scalable and portable to large corporations as well as the public sector, builds on 15 years of cooperation with ESA, and creates room for further projects and expansion of the Mistral model portfolio.
