The European Union (EU) and Latin America and the Caribbean (LAC) are deepening their collaboration to foster the development, deployment, and responsible use of artificial intelligence (AI), grounded in a shared human-centric approach.
As part of this effort, the EU-LAC Digital Alliance convened the third edition of the High-Level Policy Dialogue on AI and Online Platform Governance, held from July 1 to 3, 2025, in São Paulo, Brazil.
Co-hosted by the EU and the Government of Brazil, the event was organised in cooperation with the International Research Centre on Artificial Intelligence (IRCAI, under the auspices of UNESCO), the German Development Agency (GIZ), the United Nations Economic Commission for Latin America and the Caribbean (ECLAC), and the Brazilian Network Information Center (Nic.br).
The Dialogue gathered over 115 participants from 31 countries, including representatives of regional organizations, academic institutions, research centres, civil society organizations, and the private sector, reflecting the multi-stakeholder nature of the EU–LAC Digital Alliance.
The opening session featured remarks by Hugo Valadares, Director for Science, Technology and Digital Innovation at Brazil's Ministry of Science, Technology and Innovation (MCTI), and Jean-Pierre Bou, Chargé d'Affaires a.i. of the EU Delegation to Brazil. Juha Heikkilä, Adviser for Artificial Intelligence (AI) in the European AI Office, delivered a keynote speech about the EU AI Act and Europe's plan to become a global leader on AI, an AI Continent.
One of the highlights of the Dialogue was the presentation of the Hamburg Declaration on the Responsible Use of AI for the Sustainable Development Goals (SDGs), which Slovenia and Brazil officially endorsed. These countries jointly invited other countries in the region to join the initiative.
Based on takeaways from previous Policy Dialogues, the discussions focused on advancing concrete joint initiatives in High-Performance Computing (HPC) infrastructures, large language models (LLMs), and data protection for AI. Participants underscored that robust HPC capacity is a foundational requirement for AI research and innovation. Europe currently operates ten supercomputers across the continent, while Brazil leads the Latin American region with eight systems listed in the global Top500 ranking. Mexico also maintains a strong HPC presence, and Argentina recently joined the list. Overall, the region's HPC ecosystem is expanding, with multiple countries contributing to regional capabilities.
Among the milestones discussed was the creation of a bi-regional HPC network, supported by an initial budget of €3 million. This effort builds on the BELLA programme, which provides high-capacity connectivity between Europe and Latin America, enabling secure digital environments for scientific collaboration, advanced simulation, big data processing, and the testing of AI models.
Within the framework of the EU–LAC Digital Alliance, the BELLA II project, led by RedCLARA, is a flagship initiative under the EU's Global Gateway strategy. It aims to strengthen the region's digital ecosystem for science, technology, education, and innovation, while expanding high-speed connectivity to as many countries in LAC as possible.
A key deliverable of BELLA II is the development of a regional HPC testbed, offering a secure and scalable environment for Latin American researchers to conduct complex simulations, run big data analysis, and validate advanced processing models. Luis Eliécer Cadenas, Executive Director of RedCLARA, presented the capabilities of the testbed, which is being implemented in collaboration with SCALAC (Advanced Computing System for Latin America and the Caribbean).
Participants also visited the Centre for AI and Machine Learning at the University of São Paulo (USP), where they explored the development of LATAM-GPT—a large language model specifically created for and in Latin America. This collaborative initiative, led by Chile's Ministry of Science and CENIA, brings together more than 30 institutions across the region. It aims to support technological sovereignty, preserve linguistic and cultural diversity, and mitigate algorithmic bias in AI applications.
In parallel, data protection authorities from Brazil, Uruguay, and Spain discussed their evolving roles in overseeing AI systems that process personal data, emphasizing the importance of safeguarding citizens' privacy and rights in the digital age.
The Dialogue also addressed the need to align and coordinate regional strategies on AI governance and digital transformation, including the EU–LAC Digital Alliance, the Montevideo Declaration led by UNESCO–CAF, and ECLAC's eLAC 2026 agenda. As a practical example of regional cooperation, Costa Rica presented the OECD's AI Toolkit for Emerging Economies, developed with support from the EU and other international partners.
The outcomes of the São Paulo Dialogue set the stage for the upcoming EU–CELAC Summit, scheduled for November 9–10, 2025, in Santa Marta, Colombia. There, both regions expect to continue advancing a shared digital agenda that promotes strategic autonomy, inclusive economic growth, and a global governance framework for AI aligned with democratic values and human rights.
This article draws on information from the European External Action Service:
EU–LAC Digital Alliance: Leading on Human-Centric Artificial Intelligence
Photo: European Union
