Federal judge holds back on Anthropic's $1.5bn author settlement

"A federal judge in San Francisco declined on Thursday to give final approval to Anthropic's proposed $1.5bn settlement with authors who say the company trained its Claude models on pirated books, asking lawyers for more detail on counsel fees and payments to the three lead plaintiffs before she would sign off."

"Judge Araceli Martínez-Olguí n, who inherited the case from Judge William Alsup after a reassignment earlier this year, presided over the fairness hearing in Bartz v. Anthropic . The class is led by thriller novelist Andrea Bartz and nonfiction writers Charles Graeber and Kirk Wallace Johnson, and the deal would resolve a suit alleging Anthropic downloaded more than seven million books from the shadow libraries LibGen and Pirate Library Mirror to train its models."

"What the judge wanted to talk about was the slice that does not go to authors. Class counsel had already trimmed its fee request from 15% to 12.5% of the fund in a proposed order filed before the hearing, plus about $3m in expenses, an $18.22m cost reserve, and $50,000 service awards for each of the three lead plaintiffs (per the Authors Alliance docket summary ). The court asked for more detail on each of those line items."

"On the headline arithmetic, the settlement covers roughly 480,000 works and pays out around $3,000 per work after fees, the largest known sum in a US copyright case. Per a claims report filed in April, claims have been registered for about 91% of the eligible works, an attorney for the class told the court that the number was now above 92% as of Thursday."

A federal judge in San Francisco declined to grant final approval to Anthropic’s proposed $1.5 billion settlement with authors alleging unauthorized training of Claude models using pirated books. The judge requested additional detail on counsel fees and payments to the three lead plaintiffs before signing off. The class is led by Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, and the case seeks to resolve claims that Anthropic downloaded more than seven million books from shadow libraries to train its models. The settlement would cover about 480,000 works and pay roughly $3,000 per work after fees, with claims registered for about 91% of eligible works and above 92% as of the hearing. The judge focused on the portion not going to authors, including reduced fee requests, expenses, a cost reserve, and service awards, while objections raised concerns about settlement size and notice effects.

Read at TNW | Anthropic