Machine-translated from Spanish. It may contain errors.
Ibero-American countries agree on an alliance to build their own AI
The 30th Ibero-American Summit in Madrid launched a voluntary alliance to share data, models and compute in Spanish, Portuguese and Indigenous languages.
The 22 countries of the Ibero-American community met on November 4 and 5 in Madrid for the 30th Ibero-American Summit, and the most visible digital outcome was the Ibero-American Artificial Intelligence Alliance.
The idea is to reduce dependence on technologies built in other regions and to work with models that better understand Spanish, Portuguese and Indigenous languages. The alliance rests on three pillars:
- Data: a federated Ibero-American Data Space, not a single database; each country keeps control over what it chooses to share.
- Models: development of models adapted to the region's languages.
- Compute: shared access to computing capacity.
Joining is voluntary. Projects already underway are cited as precedents: LATAM-GPT (Chile), SoberanIA (Brazil), ALIA (Spain) and AMALIA (Portugal). The Dominican Republic also proposed an Ibero-American AI center, but no decision has been made yet on where it would be based.
Why it matters for builders
For now, it's a political agreement, with no public budget or timeline. What's worth watching is whether the data space and shared compute end up being accessible to startups, and not just to governments and universities.
It was also reported by Prensa Latina, which presents it as part of the Madrid Declaration.
If your product relies on models trained in English, what would you gain from open regional models and data?
Was this useful? Get stories like this every Monday.
Five stories, one question for builders and three new profiles.