The Eclipse Foundation unveiled the Sovereign AI Foundation on Wednesday, a vendor-neutral body designed to address the growing challenge of AI vendor lock-in. According to Mike Milinkovich, executive director of the Eclipse Foundation, the initiative brings together 17 founding organizations to help enterprises avoid becoming overly dependent on a single AI provider.

Switching between AI providers currently carries substantial costs. Organizations that have deeply integrated a specific vendor's AI systems often face expensive rebuilding of workloads, data migration challenges, and the complexity of untangling integrations. The Sovereign AI Foundation aims to create pathways that preserve organizational choice and control over AI infrastructure.

The costs of deep AI dependencies

Milinkovich explained to The New Stack that the severity of lock-in risks correlates directly with how deeply AI systems penetrate an organization's operations. He noted that regardless of how embedded these systems become, the financial and operational consequences remain consistent across enterprises.

The deeper the dependency, the greater the potential risk. But however embedded the underlying systems are, the costs show up in predictable places.

Mike Milinkovich, Eclipse Foundation

These predictable costs include inflated pricing that discourages switching, the burden of reconstructing integrations, data relocation efforts, retroactive system documentation, compliance complications, and service interruptions. The Sovereign AI Foundation plans to document these real-world experiences and develop stronger evidence about the true scope of such risks.

Preserving choice and control

Milinkovich emphasized that AI cannot fulfill its potential if organizations depend on a narrow set of vendors for critical models, platforms, and infrastructure. As AI becomes embedded in daily operations, this dependence can obscure where data travels, which models process it, and how readily an organization can transition when circumstances change.

The foundation's vision centers on delivering open-source and open-weight solutions that enable freedom of action. When asked whether geopolitical considerations—particularly regarding open-weight models from China—influenced the initiative, Milinkovich clarified that the concern transcends geography.

This is not about China or any single provider. The same questions apply to AI technologies from the United States, Europe, Asia, or anywhere else. Closed models can limit visibility, while access to model weights alone does not provide the transparency, control, or the independent governance that organizations may need.

Mike Milinkovich

The founding coalition

The Sovereign AI Foundation launches with participation from 17 organizations spanning technology companies, industrial enterprises, research institutions, and open-source contributors. The roster includes France's CEA LIST, Thales, and Kentyou; Germany's EclipseSource, Open Elements, TypeFox, Vector Informatik, and Bosch; Italy's Engineering Ingegneria Informatica and Eurotech; Sweden's Ericsson; India's Infosys; Belgium's KU Leuven; Japan's Renesas Electronics; Malaysia's The IO Foundation; North Carolina's Red Hat; and the University of York in the UK.

The foundation concentrates its efforts across five strategic areas:

  • Mapping real-world use cases to build shared understanding of how members deploy, evaluate, and test open-source AI
  • Assessing technologies and trends across models, evaluation frameworks, inference infrastructure, and regulatory developments
  • Developing practical resources including whitepapers, blueprints, landscape analyses, and reference materials
  • Building a peer community that connects organizations at various stages of open-source AI adoption
  • Translating research into action by converting outcomes from Eclipse Foundation research initiatives, including EU-funded projects, into actionable member insights

Addressing efficiency, security, and resilience simultaneously

When asked whether the Sovereign AI Foundation prioritizes data control, cybersecurity, or lock-in mitigation, Milinkovich indicated that all three concerns reinforce each other. Rebuilding workloads to accommodate a different model represents an efficiency problem. Insufficient visibility into AI systems creates security vulnerabilities. And excessive dependence on a single provider becomes a resilience and lock-in issue.

The need here is clearest wherever AI is used in systems with long lifecycles, demanding safety and security requirements, or sensitive data.

Mike Milinkovich

Milinkovich noted that specific requirements vary by sector. A manufacturing operation might require AI systems that function reliably with intermittent connectivity, while a government agency needs to document and justify AI-assisted decisions for accountability purposes. Industries with the most pressing sovereignty requirements include automotive, manufacturing, telecommunications, aerospace and defense, energy, and the public sector.

Global expansion anticipated

While the founding membership reflects a significant European presence—with companies such as Bosch, Vector, Renesas, Ericsson, and Thales prioritizing sovereignty—Milinkovich confirmed that dependence on critical AI technologies represents a worldwide concern. The foundation anticipates growing interest from organizations beyond Europe in the coming weeks and months.

Several use cases already demonstrate the foundation's practical value. EclipseSource collaborates with industrial customers on AI-enabled development tools like Eclipse Theia AI, where teams require flexibility in model selection and deployment options. The Open VSX Registry delivers vendor-neutral access to thousands of extensions for open-source and AI-enabled development tools. Eurotech applies Eclipse-governed projects to deploy AI at the edge, where systems must interact with physical infrastructure while maintaining security and reliability over extended periods. Eclipse LMOS, originally created at Deutsche Telekom, powers production multi-agent systems at scale and supports multiple global deployments.

Milinkovich emphasized that modularity represents the common thread across these implementations: maintaining sufficient separation between models, infrastructure, and applications so that changes to one component do not necessitate rebuilding the entire system. The foundation's mission includes capturing these successful patterns and converting them into guidance that other organizations can apply.

Beyond the 17 founding participants, additional organizations are actively joining the Sovereign AI Foundation. The initiative welcomes participation from eligible Eclipse Foundation member organizations.