From the Menu to the Table
When President Gabriel Boric said it at the launch of LatamGPT — "if we are not at the table, we are on the menu" — he was not being rhetorical. He was describing, with surgical precision, Latin America's geopolitical reality in the algorithmic era. Global language models are trained with barely a 2%…

When President Gabriel Boric spoke those words at the launch of LatamGPT — «if we are not at the table, we are on the menu» — he was not being rhetorical. He was describing, with surgical precision, the geopolitical reality of Latin America in the algorithmic era.
Global language models are trained on barely 2% to 3% Latin American data. That means that when ChatGPT or Gemini talk about us, they do so from the perspective of the diner reading the menu, not the cook who knows the ingredients. Our stories, our linguistic nuances, our collective memory are reduced to a statistical footnote in corpora trained predominantly in English.
That is precisely why the result of LatamGPT's cultural benchmark deserves a reading that goes beyond immediate celebration. The Dominican Republic took first place with a 16.2%, ahead of Ecuador, Colombia, Mexico, Peru and even Chile, the country leading the model's development. This is neither an accident nor a stroke of luck. It is the direct result of a strategic decision we made from the private sector: to contribute more than 42,000 journalistic records from the historical archive of CDN Canal 37 to the training of LatamGPT, through the alliance between CDN, Tabuga and CENIA.
In the final stretch, Tabuga managed to collect more than 300 historical, cultural and geographic contexts of the country. With that contribution, the Dominican Republic became the first Caribbean country to contribute data from the private sector to a massive regional artificial intelligence model.
This is where the story becomes more complex and, frankly, more revealing.
That cultural leadership in the benchmark did not come from the State. It did not emerge from a coordinated public policy or a national plan to digitize heritage. It emerged from private initiative, from a media company and a technology company that understood that a country's cultural identity is defended not only through speeches and narratives, but with structured, curated data available so that algorithms can learn who we are.
While the country has fallen in government innovation indices — dropping to position 97 in the Global Innovation Index and to ninth in the ILIA 2025, with scores of barely 47.43 in enabling factors — the private sector showed that it is possible to take a seat at the regional table when there is vision and the will to execute.
This contrast is worth pointing out: what we have been reporting from 2022 in articles such as «Artificial Intelligence in the Dominican Republic: Challenges and Opportunities for Regional Leadership» materializes here with uncomfortable clarity: the gap between the country's digital rhetoric and its real capacity for technological production. We have a ratio of 30 to 1 between AI consumption (more than 300 million dollars) and local investment in AI development (approximately 10 million). LatamGPT holds up a mirror to us.
We can lead a cultural benchmark with historical journalistic content, but we lack the computing infrastructure, the regulatory frameworks and the institutional coordination to sustain that leadership systematically.
What LatamGPT validates is something I have insisted on calling cognitive sovereignty. It is not just about having a language model of our own — which is already a monumental achievement for the region. It is about making sure that the artificial intelligence that describes us, analyzes us and makes decisions about us does so with real knowledge of our context.
When a global model «hallucinates» about our history, our literature or our geography, it is not making a minor technical error; it is perpetuating a form of cultural invisibility with concrete economic and political consequences. The data we contributed from CDN — news, chronicles, coverage documenting decades of Dominican life — is not simply content. It is memory. And memory, in the algorithmic era, is power.
That said, the cultural benchmark is a starting point, not a finish line. The main challenge we face is not technical but structural: the digitization and opening of public data. Libraries, national archives, cultural records, documentary heritage — an enormous amount of content that defines our identity remains in formats that artificial intelligence systems cannot access. We cannot build cognitive sovereignty on a fragmented and incomplete database.
If the private sector has already shown that it is possible to contribute more than 42,000 records, imagine what would happen if public institutions took on the same responsibility with the cultural and historical heritage they hold.
By being open source and auditable, the LatamGPT model offers something the closed models of the global north cannot: the possibility for universities, developers and startups in the region to build on a foundation that respects local identity without paying what I call a «cognitive toll» to the big tech companies.
Training took place on AWS infrastructure but was coordinated by CENIA, with the participation of 15 countries and more than 100 professionals — many of them volunteers. This confirms a thesis we have consistently defended: we do not need to build everything on our own. The path is federated regional collaboration, Open Innovation applied at a continental scale.
For those of us who closely follow the region's technological geopolitics, LatamGPT represents the difference between being an ingredient on the menu — data extracted, processed and monetized by others — and sitting down to design the dish. The Dominican Republic took a first step from the private sector. Now comes the uncomfortable but necessary question: when does the State take a seat at the table?
Building technological sovereignty does not depend solely on launching our own models, a misguided strategy of the local authorities. Collaboration is the act of courage that 13 countries demonstrated over more than 12 months. It depends on ensuring that our information is digitized, organized and available. It depends on training local talent that does not merely use these tools but adapts and improves them. It depends on real digital co-responsibility among the public sector, the private sector, academia and civil society.
LatamGPT's cultural benchmark tells us something powerful: when we contribute our voice, artificial intelligence listens. The question is whether we will have the collective vision to ensure that this voice does not remain a single result, but becomes a sustained strategy for positioning ourselves in the knowledge economy.
In the age of Artificial Intelligence, the budget is not spent on articulation but on compute. He who has ears to hear, let him hear.