Anthropor agrees to a $ 1.5 billion settlement to download pirated books to train artificial intelligence

Antarubor agreed to pay $ 1.5 billion to settle a lawsuit filed by authors and publishers regarding its use of millions of books protected by copyrights to train models on Ai Chatbot Claud, according to a legal report Posted online.

Federal judge It was found in June Anthropor’s use of 7 million pirated books has been protected in light of the fair use, but the digital works in a “central library” violated the law of copyright. The judge has spent that the company’s executives know that they are downloading pirate business, and a trial is scheduled to be held for the month of December.

The settlement, which was submitted to a federal judge on Friday, needs final approval but will pay $ 3000 for each book for hundreds of thousands of authors, according to what he said New York Times. The $ 1.5 billion settlement will be the largest compensation in the history of the copyright law in the United States, although the amount paid per work was mostly higher. For example, in 2012, a woman in Minnesota paid about $ 9,000 for each download song, a number that was dropped after it was initially requested to pay $ 60,000 for each song.

In a statement by Gizmodo on Friday, Anthropor described the former rule from June as being involved in a fair use by training models with millions of books.

“In June, the District Court issued a prominent ruling on the Acts of Development and Publishing Rights to Amnesty International, and found that the Anthropor approach to training artificial intelligence models constitutes fair use,” said Aparna Sridhar, Deputy General Adviser in Anthropor, said in an e -mail statement.

“Today’s settlement, if approved, will solve the remaining old claims for the plaintiffs. We are still committed to the development of safe systems of artificial intelligence that helps people and organizations extend their capabilities, enhance scientific discovery, and solve complex problems. ”

According to the legal file, Antarubor says the payments will come out In four segments linked to the monuments based on the court. The first batch will be 300 million dollars within five days after the approval of the initial court on the settlement, and another $ 300 million within five days of the final approval order. Then it will be $ 450 million, with interest, within 12 months of the initial order. Finally, 450 million dollars during the year after that.

Anthropier, who was recently estimated 183 billion dollarsShe still faces lawsuits from companies such as companies I respondedWhich concluded a deal in Early 2024 To allow Google to train artificial intelligence models on the content of the basic system. Authors still have active claims against other major technology companies such as Openai, Microsoft and Meta.

The sentence issued by June explained that the anthropologist for artificial intelligence models with copyright books are considered fair use under the law of copyright in the United States because in theory, a person can read and simulate “all modern classic classics”, which will be protected:

… She did not reproduce the creative elements of a specific work, not even a single style expressing the author … But if someone reads all modern classics because of his exceptional expression and preserved it, then simulates a mixture of their best writing, does that violate the law of copyright? Of course not.

The verdict said: “Like any reader who aspires to be a writer, LLMS of Antarbur has been trained on the works not to race forward and repeat or replace it – but to convert a solid angle and create something different.”

Under this legal theory, all the companies needed to do was to buy every pirated book to train their models legally, which certainly costs less than $ 3,000 per book. But as the New York Times notes, this settlement will not determine any legal precedent that can determine future cases because it will not be tried.