Santiago (Chile) (AFP) – Chile launched Latam-GPT on Tuesday, an initiative aimed at providing Latin America with its own artificial intelligence model to limit certain biases and stereotypes, in a sector dominated by the United States.

Published on: 02/10/2026 - 04:59 Modified on: 02/10/2026 - 17:43

Coordinated by the Andean Country's National Center for Artificial Intelligence (Cenia), a private entity financed with public funds, the project brings together institutions from at least 15 countries in the region.

Partners include universities, foundations, libraries, government entities and civil society organizations from Uruguay, Brazil, Colombia, Mexico, Peru, Ecuador, Argentina and Chile.

“Thanks to Latam-GPT, we are positioning the region as an active and sovereign actor in the economy of the future. We are at the table, not on the menu,” declared President Gabriel Boric during the presentation of the project on the public channel Television Nacional.

Despite its name, Latam-GPT is not a conversational agent intended for the general public, but a linguistic model serving as a basis for regional applications, accessible as open source and free of charge for companies and public institutions.

Latam-GPT aims to break certain “prejudices” and avoid a uniform representation of Latin America in the world, Science Minister Aldo Valle explained to AFP.

The region "cannot be just a passive user or receiver of artificial intelligence systems. This could lead to the loss of a large part of our traditions," he added.

When requesting to represent "a Chilean" to ChatGPT, the agent generated an image of a person in traditional indigenous clothing in front of the Andes mountain range, illustrating the stereotypes that the project's promoters say they want to reduce.

According to them, this type of representation is explained in particular by an under-representation of Latin American data in certain artificial intelligence systems.

President Boric thus took the example of the siege of Calais, in France, begun in 1346 during the Hundred Years' War, for which ChatGPT provides "very developed information". “But if you ask for information on the Siege of Chillán, one of the most important battles of the Chilean War of Independence, there is much less,” he assured.

Boundaries

If the development of large AI models remains largely dominated by players based in the United States, China and Europe, regional initiatives, including Latam-GPT, but also SEA-LION in Southeast Asia or UlizaLlama in Africa, are trying to adapt these technologies to specific cultural contexts.

Trained on more than eight terabytes of data, a volume equivalent to millions of books, Latam-GPT was developed with funding of $550,000, mainly from the Development Bank of Latin America (CAF) and own resources.

For some experts, this gap in resources with major international projects constitutes a major limitation.

“There is no possibility that (Latam-GPT) can compete with large AI models,” Alejandro Barros, professor at the industrial engineering department at the University of Chile, told AFP.

For the academic, major international projects have “hundreds of millions of dollars, or in some cases billions of dollars, for infrastructure”.

Despite these limitations, those responsible for the project highlight its potential applications. According to the director of Cenia, Alvaro Soto, Latam-GPT could in particular serve as a basis for tools adapted to the health or public services sectors.

"Slang"

“The data to provide a solution is not found elsewhere in the world, it is here,” he explains.

One of the first companies to use Latam-GPT will be the Chilean company Digevo, which plans to develop conversational robots specialized in customer service, notably for airlines.

Client companies “are very interested in their users being able to express themselves and receive answers in their language,” Roberto Musso, director of the company, told AFP. According to him, Latam-GPT could better recognize "slang (and) idioms...".

Currently, the data used to train the model is mainly in Spanish and Portuguese, but the project plans to gradually integrate indigenous languages.

© 2026 AFP

The content you are trying to access does not exist or is no longer available.