We review meta in AI training copyright cases

A US judge on Wednesday handed over the author who accused the tech giant of violating copyright laws by training Lama artificial intelligence without permission.

San Francisco District Court Judge Vince Chhabria ruled that Meta was “transformative” enough to constitute “fair use” under copyright law, using her work to train AI models.

However, the authors have come with the warning that by training powerful generator AI with copyrighted works, they can pitch a winning argument that technology companies are creating tools that allow a sea of ​​users to compete with them in the literary market.

“In any case, it is difficult to imagine that transformative (generic AI) training would be fair to develop tools to make billions or trillions of dollars using copyrighted books while still allowing the creation of streams of competing works that could seriously harm the market for those books.”

Training large-scale language models that power generation AI requires a huge amount of data.

Musicians, authors, visual artists and news publications sues a variety of AI companies that use their data without permission or payment.

AI companies generally defend their practices by asserting fair use, claiming that training AI on large datasets fundamentally transforms the original content and is necessary for innovation.

“We are grateful for the decision from the court today,” a Meta spokesperson said in response to an AFP investigation.

“The open source AI model is to promote transformative innovation, productivity and creativity among individuals and businesses, and fair use of copyright material is an important legal framework for building this transformative technology.”

According to court documents, the group of authors sued Meta in the previous case of Chhabria for downloading pirated copies of the work and using them to train open source llama-generating AIs.

The book caught up in the suit includes Sarah Silverman’s comic memoir, The Bedwetter, and Junot Diaz’s Pulitzer Award-winning novel, The Brief Wondrous Life of Oscar Wao.

“The decision does not represent the proposition that Meta uses copyrighted material to train language models,” the judge said.

“It only represents the proposition that these plaintiffs had made the wrong argument and were unable to develop a record in favor of the right thing.”

Market harm?

On Monday, another federal judge in San Francisco sided with the AI ​​company humanity in training models on copyrighted books without the author’s permission.

District Court Judge William Alsp held that under the doctrine of “fair use” of US copyright law, training for Claude AI models, including purchases or pirated copies, was permitted.

“Using problematic books to train Claude and its precursors was a very transformative and fair use,” Alsup wrote in his decision.

“The technology in question was one of the most transformative things that many of us could see in our lifetimes,” Alsup added to his decision, comparing how humans learn by training AI and reading books.

The ruling comes from a class action lawsuit filed by authors Andrea Burtz, Charles Graeber and Kirk Wallace Johnson.

Alsup rejected humanity’s bid for blanket protection and determined that the company’s practice of downloading millions of pirated copies to build a permanent digital library was not justified by fair use protection.

©2025 AFP