Nonetheless, the judge highlighted some vulnerabilities in the broader context of Big Tech’s AI initiatives and Meta’s rationale that it was fair use. Judge Chhabria clarified, “This ruling does not declare that Meta’s use of copyrighted materials for training language models is legal.”

“It merely indicates that these plaintiffs presented flawed arguments and did not develop a suitable record to support a valid one.” This decision comes on the heels of Anthropic securing a fair use triumph, with a different federal judge determining that using legally purchased books to train models constitutes fair use.

Judge Chhabria says that two of the authors’ arguments about fair use were “clear losers:” the ability for Meta’s Llama AI to reproduce snippets of text from their books and that Meta using their works to train its AI models without permission diluted their ability to license their works for training. “Llama is not capable of generating enough text from the plaintiffs’ books to matter, and the plaintiffs are not entitled to the market for licensing their works as AI training data,” the judge wrote.

The plaintiffs didn’t do enough for a “potentially winning argument” that Meta’s copying would create “a product that will likely flood the market with similar works, causing market dilution,” according to Judge Chhabria. He also discussed the Anthropic ruling, saying that Judge William Alsup brushed aside concerns about the harm generative AI could “inflict on the market for the works it gets trained on.”