Meta Wins Ai Copyright Case, Judge welcomes others to bring lawsuits

Meta On Wednesday, a group of 13 authors in a major copyright case, related to the model of the artificial intelligence of the company Llama, but the judge made it clear that his ruling was limited to the case.

US District Judge Vinc Chhabri on the Meta argument is that the use of the company’s books to prepare its large language models or LLM is protected in accordance with the doctrine of fair use of the US copyright law.

The lawyers representing the plaintiffs, including Sarah Silverman and Ta Nehisi Couts, claimed that Meta violated the copyright law because the company did not seek the authors’ permission to use its books for the AI ​​company.

In particular, CHABRA said that “as a rule, to illegally copy protected work without permission,” but in this case the plaintiffs did not submit a convincing argument that the use of Meta books to prepare Llama has caused “market damage”. Chhabria wrote that the plaintiffs made two imperfect arguments in their business.

“In this record, Meta defeated the plaintiff’s argument that his copying caused or threatens significant damage to the market,” Chhabri said. “This conclusion can be in considerable tension with reality.”

Practice meta “copy work for a transformation goal” are defended by the doctrine of fair use, the judge writes.

“We appreciate the court’s decision today,” said Meta press -secretary. “AI models with open source is the power of transformation innovations, productivity and creativity for people and companies, and the fair use of copyright materials is a vital legal basis for the creation of this transformation technology.”

Although there may be real arguments that the practice of Meta data has a negative impact on the book market, the plaintiffs did not properly do their job, the judge writes.

The lawyers who submit the plaintiffs did not respond to the request for comment.

However, CHABRA noted some shortcomings in the defense of Meta, including the idea that “public interest” would be “poorly deserved” if companies and other enterprises were forbidden to “use author texts as training data without paying it”.

“The meta seems to mean that such a ruling stop the development of LLMS and other AI generative technologies in its tracks,” Chhabri wrote. “It’s nonsense.”

The judge left the door open to other authors to bring similar copyright lawsuits related to the purpose, saying that “the consequences of this resolution are limited in the large scheme of things.”

“This is not a cool action, so the ruling affects only the rights of these thirteen authors – not countless others whose works are used to teach their models,” he wrote. “And, as it should now, this ruling does not indicate that the use of meta -author materials to prepare their language models is legal.”

In addition, CHABRA noted that there is still a separate claim of the plaintiffs who claim that Meta “perhaps illegally distributed its work (through the torrent).”

Used to be this week as well Identified the federal judge which AnthropUsing books for workout You have a model Claude was also “transformational” thus satisfying the doctrine of fair use. However, this judge said the anthropic should face the allegations that he has loaded millions of pirated books to prepare his AI systems. “

“Later, the anthropic bought a copy of the book she had not previously stole on the Internet, not releasing it for theft, but it could affect the degree of legislative losses,” the judge wrote.

Watch: Meta pushes back to the WhatsApp ban on the device used by the House of Representatives.