Meta Defeats Junot Diaz, et al., in Copyright Case
A judge has dismissed a lawsuit filed by 13 authors against Meta for using copyrighted works to train AI, without declaring the practice legal.
Photo by Christopher Peterson via Wikimedia Commons, CC BY-SA 3.0
A federal judge in California has ruled that using copyrighted materials to train natural language models (NLMs), such as ChatGPT, is not necessarily illegal. On Wednesday, June 25, San Francisco Judge Vince Chhabria dismissed the case brought by 13 authors, including Junot Diaz, against Meta Platforms, awarding the company a victory.
The authors, including the award-winning Dominican writer Junot Díaz, claimed that Meta used their catalogs of published works to train their LLM, “Llama.” The authors argued that the LLM competed unfairly with the authors’ work, as it is capable of producing excerpts from the original novels. Meta admitted that they trained Llama with copyrighted content. However, Meta argued that their actions fell below the “fair use” threshold and were not illegal.
Judge Chhabria agreed that Meta Platforms was correct. However, he emphasized that his decision to dismiss the case rests primarily on the fact that the authors failed to present a proper argument to defend their position:
“This decision does not imply that Meta’s use of copyrighted materials to train its language models is legal. It only implies that these plaintiffs presented the wrong arguments and failed to build a case that supported the correct argument.”
Fair Use and Copyright
The authors' main argument was based on the idea that Llama was trained using the plaintiffs' (copyrighted) material, and the plaintiffs received no compensation for the use of their original work. The authors argue that this violates their copyright, as their work is being reproduced and copied without their consent or compensation.
In the United States, when an author establishes "copyright" over a work, means that any person or organization that wants to use the work in any way must seek the author's consent. However, if someone alters the work in a "fundamentally transformative" way, a concept called the “fair use”.
“Fair use” is a legal doctrine that permits the use of copyrighted works. “Fair use” does not require the consent or compensation of the copyright holder. The legal system established this doctrine primarily to permit activities such as criticism, education, and satirical entertainment.
Meta used the concept of “fair use” to justify its actions. They argued that the process Llama uses to generate new content, even if it is based on copyrighted work, is fundamentally transformative. Judge Chhabria conceded that “fair use” applies in this case.
Meta-e reaction Implications for the Future
Judge Chhabria was at pains to emphasize the fact that this decision applies specifically to the case of the 13 authors, and not to others. The judge explained that if the plaintiffs had argued that Meta's way of obtaining their work was illegal, they would have been more successful. Meta admitted that it used pirated sources to acquire the content. In the decision the judge produced, he urged other authors to also file lawsuits against the companies that produce LLMs, but to do so properly.
Meta was very pleased with this decision. This is the second case in which an Artificial Intelligence (AI) company has won against a group of plaintiffs. On Monday, another judge declared that Anthropic had not violated the copyright in training its LLM “Claude.”
As AI continues to evolve explosively, artists are demanding compensation for the crucial role their work plays in creating these models. ChatGPT, Claude, and Llama are predictive models that assume the correct answer by analyzing billions of lyrics written by millions of authors. What kind of reward could be commensurate with such a diverse effort?
For more stories like this, follow More Latin.
Sources:
English 
Spanish 