Yesterday, Anthropic was found not guilty of infringing the copyright of several authors' works to train their artificial intelligence (AI) model. The judge agreed that using copyrighted works to train AI falls under the category of fair use according to existing law. This morning, Meta was also found not guilty because the judge again said it was fair use.
Meta was sued by 13 authors, one of whom is actress Sarah Silverman. Although found not guilty of infringing on the authors' copyright, judge Vince Chhabria said his decision was because the plaintiffs failed to prove their case well. For example, they failed to prove that a Llama trained using the authors' works could cause market disruption because the AI could produce its own work.
Meta and Anthropic's success in dismissing the lawsuit involving the authors' chip rights will have a major impact. Although not guilty of training AI using copyrighted works without permission, both companies are now facing a case of downloading these works using Torrent which is clearly in violation of United States law. As reported last February, Meta downloaded at least 81.7 TB of pirated works through the pirated library Library Genesis (LibGen).