The World Artificial Intelligence Conference (WAIC) in Shanghai closed on July 20, confirming that the World AI Cooperation Organization (WAICO), the AI cooperation organization founded by China days earlier, is already operating with 29 member countries, among them Brazil, Cuba, Venezuela and Nicaragua, which will not be required to align with the European AI Act, the OECD principles or the G7 Hiroshima Process. The picture the week leaves behind, according to coverage of the closing, is one of two governance orders that no longer merely compete but are starting to become technically incompatible with each other.
It confirms something we had been following: on one side, the bloc that Washington is considering formalizing around Demis Hassabis’s proposal for a frontier model regulator modeled on the US financial supervisor; on the other, this Beijing-led club. Neither included Latin America in its original design, but both are courting it now that it is already built. At the closing, Huawei took the opportunity to present its Atlas 950 SuperPoD, a compute system that, according to the company, matches 6.7 times the capacity of Nvidia’s NVL144 without using components of US origin, and Xi Jinping offered 5,000 AI training slots for developing countries. The underlying message is that infrastructure, and not just the rules, is also splitting in two.
Also today
- A federal court approves the $1.5 billion copyright settlement between Anthropic and authors — it is the largest AI-related copyright settlement in history, about $3,000 per work across nearly 500,000 works; no Latin American publisher was included in the plaintiff class.
- The UK’s AI Security Institute finds the cyber gap between open and closed models has shrunk to 4-7 months — open-weights models such as GLM-5.2 and DeepSeek V4-Pro nearly match closed ones on offensive tasks, at a fraction of the cost.
- A Chilean study finds that large language models underrepresent Chile’s population compared with that of the United States — researchers from the U. de Chile, the IMFD and GobLab UAI audit the bias using the CEP survey as a benchmark.
- One AI attacked Hugging Face and another open-weights AI did the forensics — the platform confirmed a breach that affected internal datasets and credentials; the safety filters of commercial models even blocked the incident response team itself.
- Chip stocks rebound after the Kimi K3 scare, but there are doubts the Chinese model will reduce demand for Nvidia — analysts warn that its architecture of splitting work among “experts” could require as much compute or more by another route.
In the region
WAICO’s operational launch redraws the map of alignments: Brazil, Cuba and Venezuela are already inside the China-led bloc, while the rest of Latin America watches without having been consulted by either of the two competing clubs. In parallel, Argentina is deepening the recentralization of its technology institutions: after the decree that removed the data and telecommunications agencies (AAIP and ENACOM) from the technical orbit of Science and Technology, the departure of the area’s secretary, Darío Genua, was confirmed, with no replacement formalized yet. And Panama announced that on July 30 it will launch its National AI Strategy, the sixth Central American country to formalize one.
Launches
- Qwen3.8-Max-Preview, from Alibaba — Alibaba’s first multimodal model above one trillion parameters (2.4 trillion, processing text, images, video and documents), shown at the Shanghai summit as a response to Kimi K3. During the preview, paid access costs just 10% of the standard price, although its weights have not been released yet.
Threads we’re following
A few days ago we covered how China formalized WAICO with 29 founding countries and Xi offered direct cooperation to CELAC, in the same week that Washington was flirting with its own standards body. Today’s chapter is that this second bloc is no longer an announcement: it operates, it has Latin American members inside, and it arrives with its own hardware in hand. The dispute is no longer about who writes the rules and has started to be about whether those rules (pre-deployment evaluation, export controls) can even coexist.
If the technical rules of both blocs end up being incompatible, is it in Latin America’s interest to play on both boards at once, or does non-alignment become unsustainable?
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.