Meta gave a major legal verdict in the AI ​​Copyright case brought by 13 authors, alleging that the company illegally trained its AI system on its work without permission. On Wednesday, Judge Vince Chhabriya Saying this ruled in favor of Meta “Summary decision on protecting its fair use for this claim deserves a violation of copying these plaintiffs for use as LLM training data.”

However, the judge also explained some weak points and meta arguments in the ecosystem of Big Tech’s AI efforts, defending their tasks as proper use. “This decision does not stand for the proposal that the use of meta copyright content is valid to train its language model,” Judge Chhabriya said.

“It only stands for the proposal that these plaintiff made incorrect arguments and failed to develop a record in support of the right.” The ruling ruling yesterday follows the victory of anthropic’s major fair won by a separate federal judge, which decided that its models were trained on copies of copies of copies legally purchased legally. Is proper use,

Judge Chhabaria says that two of the authors about the fair use “clearly lost:” Meta’s Lama AI had the ability to reproduce the text snipites from his books and Meta reduced the ability to license his tasks for training using his AI model. The judge wrote, “Lama is not able to generate adequate lessons in the case with the books of Wadi, and the plaintiff AI is not entitled to the market to license his tasks as training data,” the judge wrote.

The plaintiff did not enough enough to “potentially winning arguments” that the meta copy “creates a product that possibly with similar functions, due to the floods in the market, due to the weakening of the market,” according to the judge sixbriya. He also discussed the anthropic decision, saying that Judge William Alsup has worried about the concerns about the loss that AI “could” provoke the market for the tasks that are trained. “