rewrite this title Meta Faces Heat: Authors Claim AI Trained on Pirated Books

The lawsuit filed on January 8 claims Mark Zuckerberg approved using a dataset from Library Genesis (LibGen), a site that provides unauthorized access to millions of books and articles.

The authors, including Christopher Golden, Richard Kadrey, and Sarah Silverman, initially filed the lawsuit in July 2023, alleging that Meta used their books to train its Llama language model without permission.

According to the lawsuit, Meta concealed critical information until a discovery deadline in December 2024, which released documents that the plaintiffs describe as “the most incriminating”.

Meta engineers stripped copyright information from the LibGen dataset to prepare it for AI training. This process, the lawsuit claims, demonstrates an intentional effort to obscure the use of pirated material.

The case details internal messages showing hesitation among engineers about downloading pirated content.

One message mentions the discomfort of using a corporate laptop to download files, but the team continued to distribute pirated content through torrenting networks by January 2024.

Court documents also highlight warnings from Meta staff about the potential fallout if the company’s use of the dataset became public. One internal memo raised concerns about media coverage that could harm Meta’s reputation with regulators.

Zuckerberg distanced himself from the decisions when questioned regarding these activities, calling such actions a “bad thing” that would raise serious concerns.

Meanwhile, James Howells recently lost his legal battle to recover a hard drive holding 8,000 Bitcoin. What did the court say? Read the full story.