Judge dismisses authors’ copyright lawsuit against Meta over AI training

San Francisco – A federal judge ranked on the side of the parental platforms of Facebook by rejecting a trial for copyright violation of a group of authors who accused the company of having stolen their works to form its artificial intelligence technology.

The decision on Wednesday of the American district district judge Vince Chhabria was the second in a week of the Federal Court of San Francisco to reject the main complaints of the copyright of the authors of books against the IA industry in rapid development.

Chhabria found that 13 authors who continued Meta “made the wrong arguments” and launched the case. But the judge also said that the decision was limited to the perpetrators in the case and does not mean that Meta use of documents protected by copyright is legal.

“This decision does not represent the proposal according to which the use by meta of documents protected by copyright to form its linguistic models is legal,” wrote Chhabria. “It only depends on the proposal that these complainants have made bad arguments and have not developed a file in support of the right one.”

The lawyers of the complainants – a group of well -known writers who includes the actress Sarah Silverman and the authors Jacqueline Woodson and Ta -Nehisi Coates – declared in a press release that “the court ruled that the companies of the AI ​​which` `fuel the work protected by copyright, are not yet raping the law. ruled in favor of Meta.

Meta said it appreciates the decision.

“Open Source models feed transformative innovations, productivity and creativity for individuals and businesses, and the fair use of copyright equipment is a vital legal framework to build this transformer technology,” said Menlo Park, in California, in a press release.

Although Meta prevailed in her request for rejection of the case, she could prove to be a pyrrhic victory. In his 40 -page decision, Chhabria has repeatedly indicated reasons to believe that Meta and other IA companies have transformed into counterfeiting on copyright in series while they form their technology in books and other works created by humans, and seemed to invite other authors to bring business to his court presented in a way that would allow them to proceed to the trial.

The judge made fun of the arguments that IA companies adhere to copyright laws old decades would slow down the progress of crucial technology at a pivotal time. “These products should generate billions, or even dollars billions for companies that develop them. If the use of copyright -protected work to train models is as necessary as companies say it, they will find a way to compensate for author holders for that.”

Monday, from the same courthouse, the American district judge William Alsup ruled that the company Ai Anthropic had not violated the law by forming his chatbot Claude on millions of books protected by copyright, but the company must always be judged for having acquired these books of pirate websites illegal instead of buying them.

But the real process of an AI system distillating itself from thousands of written works to be able to produce its own text passages qualified as “fair use” under the American copyright law because it was “typically transforming”, wrote Alsup.

In the Meta case, the authors had argued in the legal files that Meta is “responsible for a violation of the massive copyright” by taking their books in the online standards of hacked works and fueling them in the Meta generative system system.

Long and distinctly written text passages – such as those found in books – are very useful for teaching the generative chatbots of AI human language models. “Meta could have and should have paid” to buy and concede to these literary works, supported the lawyers of the authors.

Meta has countered in court documents that the American copyright law “allows the unauthorized copy of a work to transform it into something new” and that the new expression generated by AI that comes out of her chatbots is fundamentally different from the books on which he was formed.

“After almost two years of disputes, there is still no evidence that someone has ever used Llama as a substitute for reading the applicants’ books, or that they could even,” said Meta lawyers.

Meta says that Llama will not release the real work he has copied, even when he was asked to do so.

“No one can use Llama to read Sarah Silverman’s description of her childhood, or the story of Junot Diaz of a growing Dominican boy in New Jersey,” wrote his lawyers.

Accused of having drawn these books from the “shadow libraries” online, Meta also argued that the methods she used had “no impact on the nature and the aim of its use” and that it would have been the same result if the company concluded rather an agreement with real libraries.

Such transactions explain how Google has built its online Google book benchmark of more than 20 million pounds, although it also fought a decade of judicial disputes before the United States Supreme Court in 2016 made it possible to support decisions of lower courts which rejected complaints in copyright.

The case of the authors against the CEO of Meta forced Mark Zuckerberg to be deposited, and has disclosed internal conversations in the company for ethics to draw from pirated databases which have long attracted control.

“The authorities have regularly closed their areas and even pursued the perpetrators,” said the lawyers’ lawyers in a judicial file. “The fact that Meta knew that taking works protected by copyright in the pirated databases could expose the company to a huge risk is out of dispute: he triggered a climb to mark Zuckerberg and other metadata for approval. Their bet should not pay.”

The complainants appointed are Jacqueline Woodson, Richard Kadrey, Andrew Sean Greer, Rachel Louise Snyder, David Henry Hwang, Ta-Nehisi Coates, Laura Lippman, Matthew Klam, Junot Diaz, Sarah Silverman, Lysa Terkeurst, Christopher Golden and Christopher Farnsworth.

Chhabria said in the decision that even if he had “no choice” to grant Meta’s summary judgment launching the case, “in the great scheme of things, the consequences of this decision are limited. It is not a collective appeal, so the decision affects only the rights of these 13 authors – not the other ruthless people whose works are used to form its models. ”

–

Technology writer AP Michael Liedtke contributed to this story.