Chile Launches Latam-GPT: Latin America's First Homegrown AI Model Takes On Cultural Bias

Latin America Gets Its Own AI Voice

Chile has officially launched Latam-GPT, the first open-source artificial intelligence model designed by and for Latin America. Developed by Chile's National Centre for Artificial Intelligence, known as CENIA, the model was trained on more than eight terabytes of regional data, equivalent to millions of books, and assembled with contributions from universities, libraries, foundations, and government bodies across eight countries including Argentina, Brazil, Colombia, Ecuador, Mexico, Peru, and Uruguay.

Tackling Cultural Bias Head-On

The project was born from a growing concern that mainstream AI systems trained primarily on US and English-language data tend to default to stereotypical depictions of Latin American culture. Chile's Science Minister Aldo Valle pointed to how major AI models often portray a typical Chilean man as someone wearing a poncho with the Andes in the background, arguing that the region cannot simply be a passive consumer of AI technology without risking the loss of its cultural traditions.

Open-Source and Community-Driven

Unlike proprietary systems from major tech companies, Latam-GPT's open architecture allows developers to customise the software for local needs. CENIA director Alvaro Soto noted that while global models do contain some Latin American data, it represents only a small proportion. The model is currently trained on Spanish and Portuguese content, with plans to incorporate Indigenous Latin American languages in future versions.

A Modest Budget With Big Ambitions

The entire project was developed on a budget of roughly five hundred and fifty thousand dollars, funded primarily by the Development Bank of Latin America and CENIA's own resources. While some experts suggest this modest investment means the model cannot compete with major AI systems head-to-head, supporters argue that cultural relevance rather than raw scale is the true goal. The first version was built using Amazon's cloud infrastructure, with future training set to move to a supercomputer at the University of Tarapaca in northern Chile.

Part of a Global Movement

Latam-GPT joins a growing wave of regional AI initiatives worldwide, including Singapore's SEA-LION model and Kenya's UlizaLlama for Swahili-speaking communities, as nations outside the dominant US-China AI corridor seek to build systems that reflect their own cultural identities.