Brazil A.I. Copyright Law Could Deter Investment, Critics Warn

The Brazil AI copyright debate has become the most contentious front in the country’s push to regulate artificial intelligence. PL 2338/2023, approved by the Senate in December 2024 and now before the Câmara dos Deputados, would require companies to pay rights holders whenever copyrighted works are used to train commercial AI systems — a stricter approach than the EU, Japan, or the United States. The bill classifies AI systems by risk level, creates a governance structure under the data protection authority ANPD, and establishes fines of up to R$50 million ($8.9 million). But it is the copyright provisions that have split opinion between those who see the rules as essential for creators and those who warn they will drive AI development offshore. The Rio Times covers Brazil financial news English and the regulatory decisions shaping Latin America’s technology landscape.

The Brazil AI Copyright Framework Explained

Under the bill, free use of copyrighted material for AI training is permitted only for non-commercial purposes by research institutions, journalists, museums, and libraries. All commercial AI development requires rights holder consent and payment under criteria of “reasonableness” and “proportionality” that the bill does not define with objective parameters. Creators gain an explicit opt-out right to prohibit use altogether, and developers must disclose summaries of training datasets.

This approach extends Brazil’s existing copyright law — one of the most restrictive in the world, where virtually all use requires prior authorization — directly into the AI domain. Critics argue it creates a regime where the cost and legal uncertainty of training AI models in Brazil could exceed that of any comparable jurisdiction. Supporters, including musicians led by singer Marisa Monte, argue it is the only way to prevent AI companies from extracting value from human creativity without compensation — making the Brazil AI copyright question ultimately a test of whether developing nations can protect their cultural output in the age of generative AI.

How Brazil Compares Internationally

The international contrast is stark. The EU’s Digital Single Market Directive permits text and data mining for commercial purposes subject only to an opt-out right — training is allowed unless the creator actively objects. Japan went further, permitting copyrighted works for AI training commercially, with remuneration required only when protected works appear in AI-generated output. In the United States, the fair use doctrine allows transformative uses without authorization, though courts have not definitively ruled on AI training. All three frameworks are more permissive than what Brazil proposes.

A coalition including the Brazilian Software Association (ABES) and agribusiness groups has urged Congress to adopt the EU’s opt-out model or remove copyright provisions entirely. Their argument: if training AI models in Brazil is more expensive and legally uncertain than elsewhere, companies will simply train abroad and deploy finished products in Brazil, leaving the country as a consumer of AI rather than a producer. This tension is amplified by the Redata program (MP 1.318/2025), which offers tax incentives for data centers. Industry groups argue there is little point building data centers if the regulatory environment makes it prohibitively expensive to use them for AI training.

The Stakes for Both Sides

The government has allocated R$23 billion ($4.1 billion) under the Brazilian AI Plan (PBIA 2024–2028) for computing infrastructure and Portuguese-language foundation models. Proponents of strong copyright protection argue this investment will be wasted if AI companies can freely appropriate creative works. They point to the bill’s regulatory sandbox provision, which allows companies to negotiate directly with rights holders, as evidence the framework is designed to be workable. Brazil’s Superior Court has historically interpreted copyright exceptions flexibly, suggesting courts may apply the “reasonableness” standard pragmatically once case law develops.

The bill remains under analysis by a special committee in the Câmara, with additional bills on AI governance submitted as recently as February 2026. The copyright provisions are expected to face significant amendment pressure before any floor vote. What emerges will signal whether Latin America’s largest economy positions itself as an AI producer or accepts the role of an AI consumer — a choice with consequences that extend far beyond the technology sector itself.