Meta wins the Copyright affair of the AI ​​deposited by Sarah Silverman and other authors

Federal judge Vince Chhabria has judge in favor of Meta on the 13 authors of the book, including Sarah Silverman, who continued the company For the formation of his broad language model on their published work without obtaining their consent. His court rendered a summary judgment in Meta, which means that the case was not deemed complete. Chhabria said Meta did not violate the law on copyright After the complainants had not shown sufficient evidence that the use of the work of the authors would do them financially.

In his decision (PDF), Chhabria admitted that in most cases it is illegal to transmit materials protected by copyright in their important language models without obtaining permission or paying copyright holders for the right to use their creations. “… By forming generative AI models with copyright-protected works, companies create something that often undermines the market for these works, and therefore considerably undermines the incentive for human beings to create old-fashioned things,” he wrote.

However, the court “must decide on cases on the basis of the evidence presented by the parties,” he said. For this particular case, the complainants argued that Meta’s actions cannot be considered as “fair use”. They said that their creations were affected by the use of Meta because the LLM, Llama of the Society, is capable of reproducing small extracts of text from their books. They also declared that by using their books for training without consent, Meta had reduced their ability to link their work for LLM training. The judge described the two arguments as “clear losers”. Llama is not able to generate enough text books to books, he said, and the authors are not entitled to the “granting market of licenses their works as training data on AI”.

Chhabria wrote that the argument that Meta copied their books to create a product that has the capacity to flood the market with similar works, causing market dilution, could have given the complainants the victory. But the complainants barely affected the argument and presented no evidence to show how Meta’s LLM production could dilute the market. Despite her decision, Chhabria said that her decision is limited: she only affects the 13 authors of the trial and “does not represent the proposal that meta use of copyright to form her linguistic models is lawful”.

Another judge, William Alsup, so recently On the side with anthropic In a collective appeal also brought by a group of authors who accused the company of having used their work protected by copyright without authorization. ALSUP, however, provided the use of writers and allowed them to bring anthropogenic to piracy.