A federal judge halted what would be the potentially largest American copyright settlement after raising serious objections over the upcoming $1.5 billion deal of Anthropic with authors whose books, they claim, were used in training the artificial intelligence company’s programs without authorization.
Northern District of California judge William Alsup didn’t mince words at a hearing this week, saying he felt “misled” by the parties and had an “uneasy feeling about hangers on with all this money on the table.” The judge did not sign off on the settlement, asking the parties to provide more information before he’s ready to move ahead.
Alsup’s primary concern is whether the agreement is really in the interests of writers or if the lawyers striking the agreement are just benefiting. He felt that members of the class often “get the shaft” after lawyers gain financial relief and no longer care about the specifics that are most important to the people they affect.
Judge Scrutinizes “Incomplete” Anthropic Settlement with Authors, Citing Major Flaws in Notification System
Settlement would bring an end to downloading millions of pirated books from Anthropic in order to train its AIs. Authors claim the company committed infringement on their copyrighted works in violation of their rights, prompting one of the original large copyright fights against the AIs like OpenAI, Meta, and Midjourney.
In the pact as proposed, writers would be paid around $3,000 per book, something attorneys say would provide a benchmark in other lawsuits against other technology giants.
The judge, though, described the agreement as “nowhere close to complete” due to the absence of crucial information, including details on what works will be included and how the potential members of the class will be informed.
The judge is frustrated with what he sees as lawyers holding back important questions until later. The current list in use includes about 465,000 pieces, but the final list remains unclear. Of more concern in Alsup’s mind is that there is no apparent system in place for notifying authors who would be entitled to payment.
“Very good notice” must be given to class members to ensure they can opt in or out of the settlement, the judge emphasized. This protection is crucial not just for authors but also for Anthropic, which wants assurance that the settlement will prevent future lawsuits over the same issues.
Authors vs. Publishers: A Legal Battle Over Book Claims
Justin Nelson, representing the authors through Susman Godfrey LLP, assured the court that attorneys “care deeply that every single proper claim gets compensation.” He expects a high claim rate given the case’s high-profile nature.
The judge’s grilling hasn’t satisfied everyone in the book world. Maria A. Pallante, the president and CEO of the Association of American Publishers, was critical of the court’s tactics, asserting they “demonstrated a lack of understanding of how the publishing industry works.”
Pallante argued the court’s vision of the procedure under the claims would be “unworkable” and would bring about “collateral litigation among authors and publishers for years to come.” The strain is typical of the thorny ownership questions in books with multiple authors and multiple publishers.
A Turning Point for AI Copyright Law
Judge Alsup untangled these complexities with the directive that every person who claims copyright ownership must opt in on an individual basis to the settlement. If only one copyright holder exercises the opt-out, no such work is included under the agreement. For ownership disputes, the judge referred parties to state court, not federal.
The judge was also displeased with what he called an “army” of attorneys in the settling process, including Authors Guild and Association of American Publishers members. He outlined that the additional attorneys will not be paid from the settling funds, with attorneys’ fees based upon how much cash actually finds its way into the pockets of the class members.
The parties now have until September 15 to submit a final list of covered works and address the judge’s concerns about the claims process. This deadline will test whether the settlement can be restructured to meet the court’s standards while still providing meaningful relief to authors.
The case represents a crucial moment for AI copyright law, as courts grapple with how to balance technological innovation with creators’ rights. The outcome could influence how other AI companies approach similar copyright challenges and shape the legal landscape for years to come.
