The Eclipse Foundation announced the Sovereign AI Foundation on 30 September 2026, bringing together 17 organisations to examine how open source projects could give them more choice over the AI systems they use. The new group will connect members’ requirements with projects already hosted by Eclipse, rather than build software of its own.
Key points
- The group is an Eclipse Foundation Industry Collaboration, not a standalone software foundation.
- Its initial participants include Ericsson, Red Hat, Robert Bosch and Thales.
- Members will assess existing projects and develop guidance; those projects retain control of their technical roadmaps.
An Eclipse collaboration with 17 participants
The Sovereign AI Foundation is an Eclipse Foundation Industry Collaboration rather than a separate software foundation, VMblog reported in an interview with Gaël Blondelle, Eclipse Foundation secretary and vice president of Community Operations. Blondelle described it as a forum where members can examine their technology dependencies, compare experience and assess open source options. Software and technical specifications will remain the responsibility of independently governed Eclipse projects, he said.
Eclipse named CEA LIST, EclipseSource, Engineering Ingegneria Informatica, Ericsson, Eurotech, Infosys, Kentyou, KU Leuven, Open Elements, Red Hat, Renesas Electronics, Robert Bosch, Thales, The IO Foundation, TypeFox, the University of York and Vector Informatik as the initial participants. The organisations span industry, research and open source development, according to the foundation.
The foundation defines AI sovereignty as an organisation’s ability to understand, choose, adapt and govern important AI systems while keeping other options available. It says that does not require isolation or complete self-sufficiency. Its concern is that dependence on models, platforms and services an organisation cannot fully inspect or change can restrict decisions about data processing, technology choices and changing providers as needs or terms evolve.
Mike Milinkovich, the Eclipse Foundation’s executive director, said the collaboration would connect existing projects with industry requirements and give members a neutral place to compare what they have learnt. That sets out the group’s proposed role: members would bring the problems they encounter in deployment to the communities that maintain relevant software, without taking over those communities’ technical decisions.
Theia, Enclave and PanEval retain separate roadmaps
Eclipse identifies several projects that members could evaluate. Eclipse Theia and Theia AI concern AI-enabled development tools. Eclipse Enclave provides isolated, policy-controlled environments in which AI coding agents can run. Eclipse PanEval addresses the evaluation of model capability, safety and cybersecurity, while Eclipse LMOS concerns systems that use multiple agents.
The distinction between the collaboration and a software project matters in practice. Under Eclipse’s proposed arrangement, an organisation could bring a requirement from its own AI deployment to the group, which would assess how relevant projects meet it and pass feedback to their communities. The foundation says each project keeps its own governance and control of its technical roadmap. A requirement can reach the people maintaining the code without giving the new group authority to rewrite it.
Changing the software used for a task at work could depend on whether its AI components can be inspected, adapted or replaced. Assessing open source projects against requirements drawn from actual deployments would make those options easier to compare, if the collaboration produces the guidance it proposes.
The Open VSX Registry is another part of the Eclipse ecosystem relevant to this work. It provides vendor-neutral infrastructure for extensions used by open source and AI-enabled development tools, and downloads are approaching one billion per month, according to Eclipse. The foundation says it is also exploring a broader AI registry for components including Model Context Protocol servers, agent skills and plugins.
Eclipse plans assessments and practical guidance
The collaboration’s planned work begins with mapping members’ uses of open source AI and examining dependencies across models, evaluation systems, infrastructure and agents. Eclipse says members will assess projects against their requirements, share experience from deployments and identify needs they have in common. The proposed resources include white papers, blueprints, analyses of available technologies and other reference material.
Another planned strand would turn findings from Eclipse research initiatives, including EU-funded projects, into material members can apply. The foundation says its existing projects will continue to develop software and specifications, while the collaboration concentrates on assessment, coordination and guidance. For companies building on those projects, that division would leave technical development with the project communities even when member organisations bring them new requirements.
Blondelle said participants in the collaboration will work under the Eclipse Foundation’s existing governance, antitrust and intellectual property policies.
