Federal judge sides with Meta in lawsuit over training AI models on copyrighted books
A federal judge chose the side of Meta on Wednesday in a lawsuit against the company by 13 book authors, including Sarah Silverman, who claimed that the company had illegally trained its AI models on their copyrighted works.
Federal Judge Vince Chhabria has one summary judgment – which means that the judge could decide the case without sending it to a jury – in favor of Meta, and found that the training of AI models of the company in this case in this case fell under the “reasonable use” doctrine of copyright legislation and was therefore legal.
The decision comes only a few days after a federal judge chose the side of anthropic in a similar court case. Together, these cases become a victory for the technical industry, which has spent years in legal fighting with media companies that claim that training AI models on copyright protected works is reasonable use.
However, these decisions are not the radical victories that were hoping for some companies – both judges noted that their affairs were limited in the scope.
Judge CHHABRIA has made it clear that this decision does not mean that all AI mode training on copyrighted works is legal, but rather that in this case the claimants “have made the wrong arguments” and did not succeed in developing sufficient evidence to support the correct one.
“This ruling does not state that the use of meta of copyright protected materials to train its language models is legally legal,” said Judge CHHABRIA in his decision. Later he said: “In cases where use is involved when meta’s is involved, it seems that the claimants will often win, at least where those cases have better developed records about the market effects of using the defendant.”
Judge CHHABRIA ruled that the use of meta of copyrighted works in this case was transforming – which means that the AI models of the company did not only reproduce the books of the authors.
Moreover, the claimants could not convince the judge that Meta violated the market for those authors through the copy of the books, which is a key factor to determine whether the copyright legislation has been violated.
“The claimants did not present meaningful evidence about market dilution,” said Judge Chhabria.
Both anthropic and the victories of Meta include training AI models on books, but there are various other active lawsuits against technology companies for training AI models on other copyright works. For example, the New York Times sues OpenAi and Microsoft for training AI models on news articles, while Disney and Universal Midjourney sue for training AI models in films and TV programs.
In his decision, Judge Chabria noted that defense defenses are highly dependent on the details of a case, and some industries can have stronger arguments of fair use than others.
“It seems that markets for certain types of works (such as news articles) may be even more vulnerable to indirect competition from AI output,” said Chabria.
