Anthropic has had to reach an agreement multimillionaire Not to go to trial for using pirate works to train AI models after a group of authors demanded the company for it. The figure that has allowed to reach an agreement is remarkable: You will have to pay 1,500 million of dollars plus interest to the plaintiffs, which is a payment of about $ 3,000 per pirated work.
According to the lawyers of the plaintiffs, it is believed that this agreement is the one with the greatest economic amount for copyright reached in US judicial history. Without a doubt, it is a notable agreement, which can sit in a precedent in this type of claims.
Of course, the agreement is still subject to the Court’s approval to be valid, with a hearing already scheduled for this on Monday, September 8. If approved, the amount that Anthropic will have to pay may even be higher, since the demand covers approximately half a million publications, but if the total of these works is superior, Anthropic will have to pay another additional $ 3,000 for each additional title.
In addition, as part of the agreement, Anthropic must destroy the original files he downloaded to train his models, in addition to all the copies he made of them. In addition, the agreement is only limited to past events, and does not give Anthropic or license or permission to training in the future. Nor does it free him from other demands that may arise from August 25 of this year.
This agreement ends a lawsuit filed in August 2024 by three autores, Andrea Bartz, Charles Graeber and Kirk Wallace Johnson, who claimed that the company had developed «A billionaire business stealing hundreds of thousands of books protected by copyright«.
Throughout this summer, a federal judge had given a minimum advantage to Anthropic, since he ruled that the company was in his right to train his AI models with books bought legally. But he opened the door to the authors, since he said that the company had to face a separate trial, because apparently he had used pirated books for training.
Last July, a federal judge of California decided that the authors could launch a collective claim representing all US authors whose works had been used to train the AI without paying for them, using pirated copies and discharged by Anthropic. The company hastened days later to reach an agreement to avoid a trial, but the terms of the agreement were unknown until recently.
Although many of the agreement may seem insufficient, what it does is sit on the protection of copyright so that other authors and companies with content and works protected by Copyright can also claim for the use of their works obtained without paying, discharged pirated from the Internet.
Meanwhile, Anthrophic and other companies dedicated to AI see the demands against him grow. Only Anthropic, for example, will have to deal with one of Reddit, who claims that the company’s bots have accessed their contents more than 100,000 times since July 2024, and has blocked them to prevent them from continuing to do so, in addition to denouncing them.
In 2023, Universal Music also demanded the startup for repeatedly violating the copyright of the letters with Copyright of the themes of its music catalog. And maybe In the future there are more authors and entities that take a step forward and demandespecially in view of the fact that Anthropic, and may be other companies in the sector, are more interested in paying strong sums of money for not complying with the law, than going to trial, where the damage to its economy can be even greater.
