This week the geopolitics of artificial intelligence gained a second protagonist. On July 16 in Shanghai, China formalized the founding of the World AI Cooperation Organization (WAICO), with 29 founding countries, among them Brazil, Cuba and Venezuela, and even UN Secretary-General António Guterres in attendance. The next day, in his first in-person keynote at the World AI Conference in Shanghai, Xi Jinping offered explicit AI cooperation to CELAC, the bloc of Latin American and Caribbean countries, and promised 5,000 training slots for developing nations.
What makes this the story of the day is not the announcement itself but the map it draws. In the same week that Demis Hassabis is promoting in Washington a standards body for frontier AI modeled on the US financial regulator, a parallel governance bloc is emerging, this one led by China, openly courting Latin America. And there lies the uncomfortable detail: neither of the two included the region in its original design, but both are seeking it out now that it is already built. For Latin American governments, the decision stops being technical and becomes strategic: which table do you sit at when none was built with you in mind?
Also today
- Moonshot AI launches Kimi K3, an open model that challenges the leaders at a fraction of the price — 2.8 trillion parameters and top-tier programming performance at $15 per million tokens, versus $50 for Western frontier models.
- An analyst warns that OpenAI could be the “Lehman Brothers” of the AI bubble — the risk of excessive concentration of capital, raised this time for a Latin American investor audience.
- Hugging Face discloses an intrusion carried out entirely by an autonomous AI agent — it would be the first publicly documented case of an end-to-end attack on infrastructure, with no human operator behind it.
- Gemini 3.5 Pro misses its third consecutive launch deadline — Google rebuilt the model from scratch after failures in tool chaining (the model’s ability to call and combine external programs).
- Tiendanube will integrate Claude for 180,000 stores in Argentina, Brazil and Mexico — on the same day that official data show only 8% of Mexican companies use AI, versus an OECD average of 20.1%.
In the region
The picture of the week is two AI governance blocs competing to define the global rules, with Latin America in the middle, having sat at the design table of neither. On one side, the financial-regulator-style standards body that Hassabis is pushing from Washington; on the other, the WAICO newly founded by China, which already counts Brazil, Cuba and Venezuela among its founding members. In parallel, several countries are advancing on their own path: Colombia is debating an AI bill in committee, still without a confirmed legislative milestone, which would place it in the region’s regulatory “second wave” alongside Brazil, Chile and Mexico. The underlying question is whether the region ends up choosing a bloc or building its own.
Launches
- Kimi K3 — an open model from Moonshot AI (2.8 trillion parameters, one-million-token context) that competes with frontier models at a fraction of the cost; its weights will be released on July 27. It lowers the barrier to accessing cutting-edge AI for developers in the region.
- Gemini Notebook (formerly NotebookLM) — Google’s AI note-taking tool is renamed and adds a “cloud computer” to run code on the sources you upload; it is already used by more than 30 million people and 600,000 organizations.
- Cosmos 3 Edge — Nvidia’s world model for industrial robotics; highly technical, but with potential for manufacturing in the region.
Threads we’re following
Just two days ago we covered here the proposal by Demis Hassabis, CEO of Google DeepMind, to create a global standards body for frontier AI, funded by the industry and backed by the United States. The objection we noted then was one of political architecture: who sits at that table. The founding of WAICO this week adds the missing chapter: now there is not one table but two, and both were set up without the region. The Latin American dilemma is no longer hypothetical.
When the rules of the game are being written in two capitals at once and neither consulted you, is it better to pick a side, wait outside, or team up with the neighbors to write your own?
About this entry. It is generated automatically from public sources, without human review before publication. It may contain errors of interpretation or summary; please check each story against its original source (the links lead there) before citing it or making decisions based on it.
Doble Click is written with Anthropic models.