Brazil’s AI Bill and Copyright: The Cost of Inaction

Brazil may remove the copyright chapter from its artificial intelligence bill just as governments and industries worldwide debate how AI systems should use protected works. The proposal, known as Bill No. 2,338/2023, is under consideration in the Chamber of Deputies after the Senate approved a substitute text in December 2024.

The current version addresses copyright in five articles, covering transparency about training data, text and data mining, creators’ ability to opt out, remuneration for the use of works, and restrictions on image, audio, voice and video content that identifies individuals. The rules would permit text and data mining only in specified circumstances and exclude commercial use.

The framework has faced criticism. Technology specialists have questioned whether some requirements, such as individually listing every work used in training, could work in practice. But the article argues that regulation would at least establish a basis for negotiation and improvement among the private sector, civil society, government and technical-scientific community. Those groups participated in 12 public hearings during the bill’s review in the Chamber.

Removing the chapter, the argument goes, would not be a neutral act. Brazil’s experience with the 2014 Internet Civil Framework offers a warning. That law left specific rules for copyright violations by application providers to future legislation, but the promised regulation never arrived. The result included years of lawsuits and frequently conflicting court decisions.

The uncertainty could be sharper with AI. Brazil’s current Copyright Law dates from 1998, before creative computing, and relies on narrowly interpreted, expressly defined limitations and exceptions. It has no equivalent to the broader U.S. fair-use doctrine. Without a dedicated framework, the article warns, disputes over training data and protected works could remain unresolved.