At a glance
- What it is: AI Policy Debates in Brazil: Struggles over Regulation, Governance and Labour
- Who: Rafael Grohmann, Edemilson Paraná, Jonas C. L. Valente, and Roseli Figaro, researchers affiliated with the AI Policy Observatory for the World of Work (AIPOWW).
- Where: Global Political Economy, Vol. 4, No. 2 (September 2025), as part of the AIPOWW symposium. doi.org/10.1332/26352257Y2025D000000037
- Type: documentary analysis. It reviews 57 documents produced between January 2019 and July 2024, following the debate through September of that year: federal and state bills, court decisions, government programs, letters, manifestos, and statements from unions and social movements.
First reading: what it does and what it finds
It helps to start with what kind of work this is, because it frames everything else. It is not an experiment or fieldwork with interviews. It is a structured reading of documents: the authors gathered 57 pieces of Brazilian public policy and read them with one underlying question, who is included and who is left out when a country decides how to govern AI. The lens is explicit: they understand these policies as part of a struggle between capital and labor, something that, they say, is almost never named. On that basis they go through three dimensions, regulation, development, and governance, and show how in all three the world of work appears only halfway or not at all.
On regulation, the picture is of a country still without a law. As of September 2024, Brazil had no national AI regulation in force; there were 15 bills in the Senate and 9 in the Chamber, and only the state of Ceará had passed its own rule. The center of the debate is Bill 2338/2023, the so-called “AI Bill,” which consolidates earlier initiatives and sets general principles. Its limit, for the authors, lies precisely in labor: the text is generic and barely touches labor rights. It came to accumulate 145 amendments, some on algorithmic management, and even so it was not voted on in 2024, held back by lobbying from big tech and from the National Confederation of Industry, which accused it of imposing a heavy regulatory apparatus.
On development, Brazil leads the region and trails far behind the world. With 163 AI-focused companies it leads Latin America, ahead of Chile (42) and Mexico (36), but that is just a fraction of the United States (5,068) or China (861). The Lula government launched its most ambitious plan in July 2024, the PBIA 2024-2028, with a projected investment of some 23 billion reais. The finding that matters to the paper is what that plan does not prioritize: labor enters almost only as workforce training, and the most concrete initiative is a platform to match job openings with candidates. The word “sovereignty” abounds; the rights of those who work do not.
Governance is where things are most alive. The authors show that the struggle is not waged only by the state and companies: unions, employer associations, NGOs, and movements step in to fight over the meaning of the policy. CUT, the largest union federation, backed the bill. The cultural sector mobilized after the 2023 Hollywood strikes, dubbing actors launched the Dublagem Viva movement against replacement by synthetic voices, and a group of researchers asked that data workers, the people who annotate and label, be covered by the law. On the business side there is no single voice: industry opposed it, while the banks’ federation, which has used AI for years, came out in favor. The paper’s message is that this governance exists and applies pressure, but it remains confined to institutional channels, letters, and hearings, without broader labor mobilization.
Second reading: from Latin America
This is one of the few Doble Lectura entries where there is no need to bring the study to the region, because it was born there. That is where its value lies: it is evidence about the largest country in Latin America, written from the inside, about a process that the Global North barely looks at.
Where you read it from matters. Brazil is the strong case, the one with the most companies, the most planning, and the most institutional muscle. If even there labor remains a blind spot, it is hard to imagine it faring better where the state apparatus is smaller. That last point is my own reading, not the paper’s: the authors studied Brazil and not the rest of the region, so I leave it as a hypothesis. But the comparison they themselves draw, with Chile and Mexico far behind in number of companies, invites the question of whether the absence of a labor perspective might be even more pronounced where there is less weight to negotiate with tech companies.
There is one nuance worth recovering just as it appears in the text. The problem they describe is not a lack of AI laws, but what those laws contain. Brazil is legislating, it has a funded plan and a civil society that pushes; and even so, algorithmic management, workplace surveillance, and job displacement remain unregulated. Translated to any country in the region, the warning is concrete: you can have a tidy, modern AI policy with investment behind it, and the world of work can still be left out of the conversation. The existence of regulation does not guarantee that it protects those who work.
The question the paper leaves, and one that applies to the whole region, is who sits at that table. The authors put it bluntly: as long as labor mobilization remains limited to formal letters, AI is likely to end up being governed without workers’ voices. The answer does not depend on the model of the moment, but on whether unions, movements, and data workers manage to move from letters to something more. On that front, Brazil is further ahead than almost all of the region, and it still has a long way to go.
The fine print
- It is a documentary analysis with a cutoff in September 2024. It portrays a process in motion, and the authors themselves warn that it does not capture later decisions, so it is worth checking how Bill 2338/2023 has progressed before citing the status of the legislative process.
- It is not a neutral piece, and that is fine: it comes from an observatory devoted to AI and the world of work, and it looks at everything through the relationship between capital and labor. That sharpens what it sees about labor rights, but it is not an even-handed assessment of every angle of Brazilian AI policy.
- It works on what the documents say, not on how they are applied. It shows what is proposed, what is omitted, and who applies pressure, not whether the rules, once passed, change anything in the day-to-day lives of those who work.
Paper keywords: AI policy, AI regulation, AI governance, labour rights, Brazil, capital-labour relations
Automated reading. This text was generated by Claude, an Anthropic model, from the original source, without line-by-line human review. It may contain errors or debatable interpretations; to check any point, see the original source.