Digital Sovereignty and Open Models

The Challenge of Sovereignty in Public Education

As Artificial Intelligence transforms education, governments and public institutions face a critical dilemma: How can these technologies be leveraged without compromising the privacy of minors’ data or relying exclusively on closed foreign corporate infrastructures?

Educational assessment (exams, rubrics, essays, learning progressions) contains the most sensitive data in the education system. Constantly sending this information to “black boxes” hosted on other continents represents an unacceptable risk to a country’s digital sovereignty.

Agnostic Architecture: Your Infrastructure, Your Rules

ColabEdu has been designed from day one with digital sovereignty as an architectural pillar. The Open Assessment Specification (OAS) is an open standard, and our assessment engine is completely agnostic to the underlying LLM (Large Language Model).

This means that public administrations can adopt ColabEdu under an On-Premise model (hosted on their own national or regional servers) using Open Weight Language Models instead of relying on third-party commercial APIs.

Integration with Public Initiatives (ALIA, LATAM-GPT)

We observe with great enthusiasm and strategic alignment the initiatives promoted by supercomputing centers and public administrations, such as the Barcelona Supercomputing Center (BSC).

Foundational projects like ALIA (for the development of models in Spanish and co-official languages) and LATAM-GPT (for Latin America) represent exactly the type of sovereign infrastructure on which ColabEdu is designed to operate.

We propose using ColabEdu’s infrastructure as the official testbed for the validation of these models in the educational field:

1. ALIA/LATAM-GPT Validation in Assessment: Demonstrate how a public foundational model can execute complex OAS standard rubrics with the same or greater pedagogical accuracy than commercial corporate models.
2. Use Cases in Co-official Languages: Utilize ColabEdu’s architecture to load regional taxonomies and rubrics (e.g., Departament d’Educació) and evaluate them entirely locally using models trained in those languages.
3. Guaranteed Privacy: Process student essays and exams on public infrastructures (such as MareNostrum) ensuring that not a single piece of data leaves institutional boundaries.

A Natural Alliance

We believe that governmental educational technology should be built upon two public pillars: Sovereign Open Models (like ALIA) and Open Data Standards (like OAS).

At ColabEdu, we aim to be the testbed and deployment platform where public AI research efforts are translated into tangible, secure, and sovereign tools for the teachers and students of our country.