Anthropic settlement faces judge scrutiny over fees and payouts

A federal judge in San Francisco, U.S. District Judge Araceli Martínez-Olguín, declined on May 14 to give final approval to a proposed $1.5 billion settlement between Anthropic and a class of authors, asking for more detailed disclosures on how the fund would be split and on counsel fee requests, according to reporting by The Next Web and NewsBytes. The settlement would cover roughly 480,000 works and, after fees, pay around $3,000 per work, per The Next Web. Class counsel told the court that claims filings covered more than 92% of eligible works, citing a claims report filed in April and an update at the hearing (The Next Web, NewsBytes). The court asked for additional detail on a revised counsel fee request reduced from 15% to 12.5%, approximately $3 million in expenses, an $18.22 million cost reserve, and $50,000 service awards for each of the three lead plaintiffs, per The Next Web.

Per court filings and reporting, the underlying litigation alleges Anthropic downloaded more than seven million books from shadow libraries including LibGen and Pirate Library Mirror to train the Claude family of models; former Judge William Alsup earlier found that training on legitimately acquired books was fair use but that retaining a central repository of pirated works raised separate infringement issues (The Next Web, NewsBytes, Mezha). Law360 reported counsel defending the fee request argued the settlement was the "creation of class counsel," a characterization used in oral argument (Law360).

Editorial analysis: This proceeding is one of the highest-profile U.S. copyright matters tied to large language model training data and, if approved as written, would be the largest known U.S. copyright settlement. Comparable high-value class settlements typically attract close judicial scrutiny over notice effectiveness, claims rates, and the share of the fund that goes to counsel versus class members; the objections reported here-complaints about insufficient relief, alleged notice gaps for non-English speakers, and allocation rules for group copyright registrations-mirror recurring pain points in mass class settlements (The Next Web, NewsBytes, Law.com).

Editorial analysis: For practitioners, the case matters because judicial handling of fee allocations, notice adequacy, and claim-administration mechanics can set procedural and practical precedents for how copyright holders and AI developers resolve training-data disputes. Separate reporting shows a contingent of authors, including novelist Dave Eggers and others, opted out and filed a new suit challenging Anthropic, underscoring that parallel litigation could continue even if a class settlement moves forward (Law.com). Observed patterns in similar cases suggest that judges will demand clear, itemized disclosures before approving any settlement that resolves large numbers of individual copyright claims.

Editorial analysis: Observers should track the court-ordered supplemental disclosures on fee calculations and service awards; the judge's eventual written order will indicate whether the court views the notice program and the allocation rules as adequate. Also watch the separate opt-out suits for distinct legal theories or requests for jury trials that could produce precedential damage rulings. Finally, monitor whether the parties revise the settlement mechanics in response to objections from authors who say the notice was late or not provided in Spanish, as reported (The Next Web, NewsBytes, Law.com).