During UNESCO’s Digital Learning Week in Paris, France, the paper ‘Digital and AI sovereignty: four evidence-informed cases’ – authored by Luciana Benotti, Beatriz Busaniche, Marcos Javier Gómez, and Margarita Trovato, members of Fundación Vía Libre – was presented.
On Tuesday, September 8, Luciana Benotti participated in the plenary session “Sustaining Education as a Common Good in the Age of AI: Towards Deliberative Governance.” During the session, she presented the paper in a Pecha Kucha format as part of the official launch of UNESCO’s compilation of policy briefs on AI and education.
In her presentation, Benotti highlighted the relevance of AI Sovereignty on the global public agenda. She argued that it should not be understood as an abstract or monolithic concept, but rather as a multidimensional framework that needs to be grounded in political and educational practice. Among her key conclusions, she emphasized:
- Commercial access limits: Mere access to commercial AI tools does not confer technology sovereignty or autonomy.
- The imperative of local evidence: Developing nations and local communities must build their own benchmarks and research to hold Big Tech corporations accountable.
- Pedagogical sovereignty as a driver of autonomy: Education must go beyond teaching instrumental AI skills; it must unveil algorithmic bias, hallucinations, and sycophancy to empower citizens to independently evaluate technology.
- Avoiding structural dependency: For both global and national educational institutions, adopting sovereign AI approaches is the only way to guarantee autonomous and genuine pedagogical decision-making.
About the paper ‘Digital and AI sovereignty: four evidence-informed cases
The paper distinguishes five dimensions of AI sovereignty in education: epistemic and pedagogical sovereignty; linguistic and cultural sovereignty; data sovereignty; infrastructural and fiscal sovereignty; and jurisdictional and technological sovereignty. and illustrates each through four cases.
The case of the Netherlands exemplifies the combination of jurisdictional and technological sovereignty. Public research and education bodies developed a national language model built with proprietary code called GPT-NL, which reduces dependence on foreign providers and keeps data within the national and European legal jurisdiction.
Another case study examines the experience of Brazil, a state that has formally elevated AI Sovereignty to an explicit objective within its national public policy. The case illustrates infrastructural, fiscal, and linguistic sovereignty.
The Masakhane network in Africa actively exercises linguistic and cultural sovereignty by developing independent benchmarks to evaluate and audit Artificial Intelligence models under their own terms and cultural standards.
Finally, the Argentina case illustrates epistemic and pedagogical sovereignty. Through a professional development course, teachers were taught the fundamentals of how AI works and how to evaluate it without relying on commercial AI, transforming educators into critical AI evaluators rather than passive users. This initiative designed a methodology focused on thoroughly comparing global benchmarks produced by tech corporations against benchmarks developed locally in the region.
To counter the risks of rapid obsolescence, unpredictable costs and cultural homogenization, the paper argues for building AI sovereignty in education by prioritizing foundational AI knowledge and local capabilities.
Global consultation on education in the age of AI
UNESCO has announced that the global consultation on the need to sustain education as a common good in the age of AI is now open. Fundación Vía Libre invites everyone to participate in this consultation to help evaluate the best conditions for integrating AI into the global educational system. Contributions can be submitted until October 15, 2026. For more information, visit UNESCO’s website.