Gabriel Boric took the stage with a clear ambition: for Latin America to stop being artificial intelligence's backyard. The Chilean president presented "LatamGPT," an AI model trained with approximately 300 billion parameters centered on data from the region. The premise is simple: if global algorithms don't understand our Spanish, our idioms, or our realities, then we need to create our own.

Boric explained that the project seeks to reduce cultural biases and affirm Latin American identity in the field of artificial intelligence. "We don't want a model trained in Silicon Valley telling us how to speak or what's relevant for us," he declared during the presentation in Santiago. The tone was combative, but the background is technical: most large language models are fine-tuned with Anglo-Saxon, European, or Asian content, leaving Latin America as a footnote.

The announcement ignited the debate about technological sovereignty. Academic and business sectors applaud the initiative; others warn the project could sink without sustained investment and regional collaboration. Protection of local data is also at stake: LatamGPT promises to use information from Latin American countries without exporting it to external servers, something that sounds good on paper but demands robust infrastructure.

Chile isn't alone. Brazil, Argentina, and Mexico have shown interest in joining the project, though financing and governance details remain unclear (data not yet confirmed by official sources from the mentioned governments). What is clear is that LatamGPT doesn't intend to compete directly with GPT-4 or Claude, but to offer an alternative that speaks with its own accent.

Time will tell if this is a strategic move or an announcement that dissolves in regional bureaucracy. For now, the bet has been placed.