Anthropic, final approval for $1.5 billion settlement in copyright lawsuit

A U.S. federal judge has definitively approved the $1.5 billion settlement agreed upon by Anthropic to close the class action lawsuit concerning the misuse of copyrighted texts. The ruling, signed by District Judge Araceli Martinez-Olguin following the preliminary opinion expressed by Judge William Alsup before his retirement, unlocks the disbursement of reimbursements destined for tens of thousands of writers and publishers for training the Claude family of language models.

The agreement encompasses approximately 500,000 literary works and stipulates a payment of about $3,000 for each book, a value that will be shared between the authors and their respective publishing houses that hold the rights. Court documents reveal that over 91% of the more than 482,000 volumes cataloged in the decision have already been claimed by their legitimate owners. The attorney for the claimants, Justin Nelson, described the settlement as the largest financial recovery ever recorded in U.S. copyright law history.

The distinction between fair use and digital piracy

The legal disputes arose from the analysis of how the company constituted its training datasets. The documents highlighted two distinct acquisition channels: scanning physical copies that were regularly purchased and mass downloading of volumes from shadow libraries and pirate portals, including Library Genesis and Pirate Library Mirror. Judge Alsup established a clear boundary in interpreting the matter, stating that the use of works for training AI models falls under fair use if the volumes were legally obtained, whereas using material sourced from piracy constitutes an unlawful act.

Anthropic's decision to settle to avoid a jury trial and the associated penalties has closed the financial dispute, but it leaves unresolved the underlying regulatory issue. As this is a settlement agreement at the district court level, the ruling does not create a binding precedent for other U.S. courts. Judges in other jurisdictions retain full independence in any ongoing litigation.

The copyright dispute in the age of artificial intelligence remains open on multiple fronts. Companies such as Meta, OpenAI, and Midjourney are still engaged in similar lawsuits. A coalition consisting of publishers like Hachette, Cengage, and Elsevier, along with associations like S.C.R.I.B.E. and writer Scott Turow, has initiated a new collective action against Google, accused of exploiting copyrighted works to train the Gemini platform.