As artificial intelligence becomes increasingly embedded in everyday products and services, the use of copyrighted material to train these systems has become one of the most disputed aspects of the technology.
This debate recently gained further attention after A.G. Sulzberger, publisher and chairman of The New York Times, criticized AI companies for using copyrighted content to train their systems without permission or payment. In particularly strong terms, he described the practice as a “brazen theft of intellectual property,” arguing that news articles, books, music, films, and other creative works have become a central part of the development of AI tools, while their authors and rights holders are often left out.
The concern is that technology companies recognize the economic value of virtually every resource involved in developing AI systems, investing heavily in computing infrastructure, energy, data storage, and specialized staff, but often treat the content used to train those systems as if it were freely available. The debate therefore also concerns how the value generated by AI should be shared with the authors and rights holders whose works contributed to its development.
For news organizations, there is an additional concern that AI-generated summaries and answers may reduce traffic to the original sources, allowing the systems to compete for the same audience as the publishers whose content they have used.
From a legal standpoint, the issue is broader than whether an AI system reproduces a protected work word for word in its output. It involves how copyrighted material is obtained, copied, stored, and processed during training, whether rights holders are informed or given the opportunity to object, and whether the resulting systems or outputs may reproduce, replace, or commercially compete with the original works.
These questions challenge traditional copyright concepts, particularly because the training process may involve large-scale use of protected material even when individual works are not immediately identifiable in the final output.
Brazilian copyright law does not yet specifically regulate the use of protected works in AI training, but Bill No. 2,338/2023 seeks to address some of these issues, including transparency and compensation for the commercial use of copyrighted material.
Until clearer rules are adopted, licensing agreements, greater transparency regarding training data, and appropriate mechanisms for compensating rights holders are likely to remain important tools for balancing technological development with copyright protection.
—
Author: Enzo Toyoda Coppola, Thaís de Kássia R. Almeida Penteado and Cesar Peduti Filho, Peduti Advogados.
Source: https://veja.abril.com.br/economia/publisher-do-new-york-times-acusa-empresas-de-ia-de-roubo-descarado-de-propriedade-intelectual/
—
“If you want to learn more about this topic, contact the author or the managing partner, Dr. Cesar Peduti Filho.”