Today Brazil’s Chamber of Deputies votes on PL 2338/2023, Latin America’s first artificial intelligence regulatory framework with full national scope. It reaches the threshold after 18 months in the legislative process: if the floor approves the text without changes, the law is one step away from the presidential signature; if it amends it, it goes back to the Senate and could be left for 2027. It is, without exaggeration, the most important AI policy event the region has ever had.
The bill follows the European model: it classifies systems by risk level, creates a National AI System and provides for fines of up to 50 million reais. The hottest point is the copyright chapter the Senate approved in 2024 and which, according to more than 50 organizations from the creative sector, rapporteur Aguinaldo Ribeiro’s report may have removed. The regional contrast is striking: in Chile, an equivalent article on protected works was rejected in mid-May. The underlying question is whether Latin America can build AI governance that is truly its own or whether it will end up adapting a mold that Europe itself simplified and postponed the same month.
Also today
- Anthropic moves toward a round of more than $30 billion — Bloomberg reported an imminent close at a valuation above $900 billion; if confirmed, it would be the most valuable private startup in the world. To put it in perspective: the region’s most ambitious sovereign compute plan is equivalent to less than 1% of this single round.
- Anthropic says Claude will remain ad-free — “Advertising incentives are incompatible with a genuinely helpful assistant,” the company argues, just as OpenAI brings ads to Brazil and Mexico. The business model defines what information millions of free users receive.
- OpenAI strengthens ChatGPT’s responses in sensitive conversations — It reports a 39% reduction in undesired responses on mental health topics compared with the previous version, and is preparing a “trusted contact” feature. An issue with weight in a region where several health ministries deploy chatbots without equivalent frameworks.
- UNESCO, UNDP and ITU open a second data governance cohort — In its first round, the free course brought together officials from 23 countries in Africa, Asia and the Arab States. Latin America was not a focus; the new edition opens in June and is an open door for ministries in the region.
In the region
It is probably the densest week of AI governance Latin America has ever experienced. While Brazil decides on PL 2338, the European Commission is moving forward with draft classifications of high-risk systems under its AI Act (the same model that inspires the Brazilian text), with public consultation open until June 23. The coincidence matters: whatever Europe ends up fine-tuning will directly affect how the future Brazilian law is interpreted and applied. And the region’s absence from the first multilateral data governance cohort, just as Brazil legislates, highlights the distance between global capacity-building processes and the legislative decisions made country by country.
Launches
- Gemini for Science (Google) — A suite of AI tools for research: scientific literature summaries, hypothesis generation through a “tournament” of ideas and computational code discovery. It includes access to more than 30 life sciences databases. Of direct interest to Latin American research groups, although the rollout is gradual.
- Gemini 3.5 Flash in general availability (Google) — Now available in the API, the Gemini app and GitHub Copilot. It promises to be four times faster than previous models in its category, at $1.50 per million input tokens: a competitive alternative for developers.
- Claude Security in public beta (Anthropic) — It scans code for vulnerabilities and proposes automatic fixes; it reports more than 2,100 flaws patched in its first three weeks. It requires an enterprise license, which aims it at security teams in companies and banks with that type of contract.
Threads we’re following
Today’s vote adds to something we had been following: the tug-of-war between two ways of regulating AI in the region. On one side, Brazil has been pushing for a year and a half a broad framework inspired by Europe; on the other, Chile rejected a key piece on protected works a few weeks ago. That this happens the same month the European Union is simplifying and delaying its own rules makes the question more urgent: does it make sense to copy a model its author is rewriting, or is there room for a path of our own?
If Brazil approves it today, the region will have a national AI law on the table for the first time. The question that remains open is how much of its own that law will be, and how much of an echo it will have in the countries that look to Brasília to decide their next steps.
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.