Judge clears Anthropic’s $1.5B copyright deal with authors

A San Francisco federal judge has given final approval to Anthropic’s $1.5 billion settlement with authors and publishers, clearing the largest known copyright recovery in U.S. history and ending the first major AI-training copyright case to reach a deal. U.S. District Judge Araceli Martinez-Olguin signed off Monday, rejecting objections that the settlement undervalued writers’ claims or improperly handled fees and exclusions. [0]

The approval lets payments move forward for rightsholders whose books were included in the class action over Claude, Anthropic’s AI chatbot. Class members filed claims covering more than 92% of roughly 480,000 works, and the settlement is expected to pay about $3,000 per work, shared among qualifying authors and publishers. courthousenews

The case drew close attention because it split two questions that are still shaping the AI industry. Former U.S. District Judge William Alsup had ruled that using books to train Claude was fair use, but he also found Anthropic could face liability for downloading and storing more than 7 million pirated books in a “central library.” That piracy issue was headed for trial before the parties reached the settlement. straitstimes

Martinez-Olguin overruled objections from writers who argued the deal was too small, gave too much to plaintiffs’ lawyers or excluded some rightsholders. She said complaints about the settlement’s size were not grounded in the risks of going to trial, where damages could have been enormous but liability was not guaranteed. straitstimes

Anthropic framed the approval as closure for a case it settled after the fair-use ruling. Deputy general counsel Aparna Sridhar said the company was pleased that more than 91% of covered authors and publishers had claimed a share of the fund. Lead plaintiffs’ lawyer Justin Nelson called it a historic settlement and said distributions should begin as promptly as possible. [0]

The settlement also requires Anthropic to destroy original files of works downloaded from pirated datasets such as Library Genesis and Pirate Library Mirror, a key noncash term for rightsholders who argued the company built valuable AI systems from unlawfully obtained books. arabnews

The approval does not settle the wider fight over AI training. Lawsuits against OpenAI, Google, Meta, Midjourney and other AI companies remain active, and because Anthropic settled rather than appealed, Alsup’s fair-use ruling does not bind other courts. courthousenews

Some authors and publishers opted out of the class deal and have filed separate lawsuits against Anthropic, leaving the company with continuing exposure even after the main settlement cleared court. nypost