An American federal court has given Meta an important victory in the ongoing legal battle of how AI companies use copyright content. On Wednesday, Judge Vince Chhabriya ruled that Meta did not break the copyright laws, when he used books to train his big language model (LLM) by 13 authors, stating that the use is eligible as “fair use”.

Judge Chhabriya wrote, “The court has no choice but to give a summary decision to Meta,” saying that the plaintiff failed to present enough evidence that the use of the company’s work caused financial loss.

A group of prominent authors, including Sarah Silverman and Ta-Nehi Kotts, brought a case against Meta in 2023, accusing the company of training its AI on its copyright books without permission. He accused Meta of using his copyright functions without permission to train his AI system. The lawsuit was among the first people of its kind, although many similar cases are now being considered by courts across the United States.

In copyright law, fair use is determined by factors such as the new work is “transformational” and whether it damages the market for the original. Judge Chhabriya mainly focused on market losses in his judgment, unlike his approach with Judge William Alsup, who ruled early this week that anthropic’s AI training was also appropriate use, but made his decision more based on the transformative nature of use.

“In fact, an important question in any case where a defendant has copied the original work of anyone without permission, will it allow people to be enacted to significantly to be attached to that kind of conduct,” Chhabaria wrote. He concluded that in this case the authors did not show that Meta’s works affected his book’s sale or earnings.

The ruling can shape future legal strategies in equal cases. Allegedly, the judge raised the issue with the idea of ​​”weakening market weakening” as a standalone cause to deny proper use. “We have not seen the final of this novel market weakening principle,” said Jacob Nota-Savar, Professor of Cardoxo Law.

Nevertheless, the judge left the door open for future claims. “In many circumstances, copying copyright-reserved tasks would be illegal to train the generative AI model without permission,” Chhabriya wrote. “Which means that companies will usually need to pay the copyright holders for the right to use their content to avoid liabilities for copyright violations.”

While Meta welcomed the verdict, the lawyers of the plaintiff expressed disappointment. Bise Schiller Flexner said in a statement, “The court ruled that AI companies who feed the copyright-reserved tasks in their models without permission are usually violating the law.” “Nevertheless, despite the undisputed record of the meta of copyright work, the court ruled in favor of the Meta. We are honorable disagree.”

Meta spokesperson Thomas Richards hit a positive tone: “Open-source AI models are strengthening transformational innovations, productivity and creativity for individuals and companies, and proper use of copyright content is an important legal framework for the manufacture of this transformational technology.”

The case comes amidst the lack of legal disputes worldwide. Earlier this week, Judge Alsup ruled that the AI ​​training practices of anthropic qualified as appropriate use, although they ordered a separate test to determine whether the company has violated copyright by storing pirated books. Meanwhile, Microsoft faces a lawsuit from the authors who claim that it uses 200,000 pirated books to train its Magtron AI model, and getty images recently demolished major copyright claims against stability AI in the UK.

All these cases come together to reflect the ongoing conflict between the creators and AI companies how creative works are used in training data. While the courts have so far bowed towards protecting the use of transformative AIs under proper use, the debate is over.