The US court backs Meta in the controversy over the use of copyrighted works to train its AI.

A California judge has dismissed the lawsuit against Meta for allegedly violating copyright laws by training its artificial intelligence (AI) with works without the creators' consent. This is the second ruling in the United States in the same week favoring AI development companies in cases where authors claim their works were used to power their models.

San Francisco District Judge Vince Chhabria ruled that the use of AI by Meta, the parent company of Facebook, WhatsApp, and Instagram, in training its AI model was sufficiently "transformative" to be considered "legitimate" under copyright law.

However, he cautions that the authors were able to make a winning argument that by training the powerful AI on copyrighted works, the tech giants are creating a tool that would allow a multitude of users to compete with them in the literary market.

“No matter how transformative (generative AI) training is, it is hard to imagine it would be fair to use copyrighted books in developing a tool to make billions or trillions of dollars by potentially enabling an unlimited flow of competing works, which could severely damage the market for such books ,” Chhabria said in his ruling.

Among the books used for training were Sarah Silverman's "The Bedwetter" and the novel "The Brief Wondrous Life of Oscar Wao" by Pulitzer Prize winner Junot Diaz, according to the lawsuit. Training generative AI language models requires enormous amounts of data. Musicians, book authors, visual artists, and news publishers have sued developers who used their works without consent or payment.

These companies defend the concept of "fair use," arguing that training AI with huge data sets fundamentally transforms the original content and is necessary for innovation. "We appreciate today's court decision," a Meta spokesperson said in response to an AFP inquiry.

"Open-source AI models are driving transformative innovation, productivity, and creativity for businesses and individuals, and the fair use of copyrighted material is an indispensable legal framework for building this technology," he added . Another court ruling on Monday dismissed claims by Anthropic that it had infringed copyright by training its AI robot with books without the creators' permission. San Francisco federal judge William Asup ruled that this practice is permissible under the "fair use" doctrine under U.S. law.