European Public AI: Reframing Sovereignty as Public Digital Infrastructure
In European Public AI – Policy Brief (January 2026), Dr. Alek Tarkowski and Dr. Felix Sieker argue that today’s AI ecosystem is being pulled toward capital-intensive, proprietary scaling that concentrates control over the full technology lifecycle—compute, data, models, platforms, and applications—in the hands of a small set of mostly non-European firms. The authors contend that Europe risks spending heavily to “enter the race” while deepening dependence on external proprietary ecosystems, unless it articulates a distinctly European “sovereign logic” for AI development—where sovereignty means resilience and sustained capability under stress, not technological autarky.
The brief frames concentration of power as a concrete systemic risk, not an abstract market outcome. From a European vantage point, only a small fraction of the most capable frontier models are European, and Europe’s broader digital dependency is substantial: a large share of digital products and infrastructure used in Europe is provided by non-EU companies, with cloud markets dominated by a handful of U.S. hyperscalers. Because cloud and generative AI are increasingly embedded in essential services—public administration, education, health care, and security-sensitive systems—control by a narrow set of foreign private actors raises issues that extend beyond economics to democratic resilience, security, environmental objectives, and fundamental rights.
Against polarized public narratives that oscillate between hype and existential alarm, the authors recommend treating AI as a “normal technology”: powerful, consequential, and infrastructural, but best governed through grounded assessment of current capabilities, limits, and dependencies. This framing underpins their core proposal: a European public AI strategy that positions AI as public digital infrastructure. In their definition, “public AI” rests on three pillars—universal, non-discriminatory access; mission-driven public goals that fill gaps markets will not; and public control through meaningful oversight, public funding or provision, and potentially participatory governance.
Critically, the brief insists that public AI must be “full-stack.” The authors describe an integrated infrastructure stack spanning compute, data, models, applications, and cross-cutting software, and argue that Europe should coordinate investments across these layers rather than treating compute expansion as an end in itself. They assess recent EU initiatives—such as the AI Continent Action Plan, Apply AI sectoral flagships, and the Data Union Strategy—as important but fragmented, warning that absent a clear public AI mandate they may function largely as subsidies that accelerate adoption of proprietary tools.
Their recommendations emphasize open-source, democratically governed “flagship” foundation models as a non-negotiable condition for sovereignty-oriented investment; a multi-client operating model for AI gigafactories suited to Europe’s demand profile; sustained funding for software “commons” and open standards; and a two-track data strategy that pairs legal clarity for training on lawfully accessible data with the creation of a European data commons stewarded by data labs. The conclusion ties these threads to “purposeful deployment”: a demand-driven approach that funds AI where benefits are evidenced and aligned to public interest goals, including the discipline to identify contexts where AI is not the appropriate solution.
Disclaimer: This article is a high-level summary of a third-party policy brief and is provided for informational purposes only. It does not constitute legal advice, investment advice, or policy counsel. Readers should consult the original publication and qualified professionals for decisions requiring reliance or implementation.