Judge Delays Anthropic’s $1.5 Billion Settlement With Authors, Microsoft Seeks Collaboration

• Judge Alsup delays approval of Anthropic’s $1.5B author settlement; new hearing set for Sept. 25th.
• Deal sets up $1.5B fund, around $3,000 per work, based on 4,65,000 works submitted on Sept. 1st.
• The court flagged Anthropic’s use of pirated books despite ruling lawful copies as fair use.
• The case unfolds as Anthropic secures $13 billion in funding, $183 billion valuation, and fresh deals.

A federal judge on Monday sharply criticized a proposed $1.5 billion settlement between AI developer Anthropic and a class of book authors who claim that the AI giant company used pirated copies of their books to train its Claude chatbot, and ordered the parties back to court with more detail.

U.S. District Judge William Alsup spent nearly an hour at the hearing warning that the proposed deal, intended to avert a December trial, leaves too many unanswered questions. He set a new hearing for Sept. 25 and ordered the parties to produce a finalized list of works and a draft claims form so he can assess whether authors will receive fair notice and compensation. “We’ll see if I can hold my nose and approve it,” Alsup said before adjourning Monday’s proceedings.

What would the settlement do?

Under the settlement filed with the court, Anthropic would put at least $1.5 billion into a non-reversionary fund to be distributed on a per-work basis, an estimated $3,000 for each covered book if the fund is allocated across roughly 500,000 works. The parties say the figure could grow if additional eligible works are identified. Plaintiffs served a draft “Works List” of about 465,000 works on Sept. 1; Alsup ordered that any revisions be exchanged by Sept. 15 and a joint list filed thereafter.

The settlement also requires Anthropic to destroy downloaded copies it obtained from shadow libraries (identified in court filings as Library Genesis and Pirate Library Mirror) and limits the release to past conduct through Aug. 25, 2025. The agreement preserves claims tied to future conduct and model outputs.

Why did the judge push back?

Alsup’s objections centered on transparency and administrative practicality. He said he needs ironclad assurances that the parties’ list of covered works is complete so Anthropic will not be blindsided by a sudden increase in the number of eligible books, and so class members do not “get the shaft” during claims administration. He set a Sept. 22 deadline to see the proposed claims form before reconsidering preliminary approval on Sept. 25.

The judge also raised concerns about the role of industry groups in the settlement process, questioning whether organizations such as the Authors Guild and the Association of American Publishers might be influencing authors’ decisions “behind the scenes.” The Authors Guild said it was “confused” by that suggestion and insisted its role has been transparent and advisory.

The legal backdrop: a mixed June ruling

The settlement follows Judge Alsup’s June 23 summary-judgment opinion, which drew a distinction that now shapes the case’s path forward: the court found Anthropic’s practice of training models on lawfully acquired books was fair use, but held that Anthropic’s downloading and retention of more than seven million pirated files for a central library could not be resolved on summary judgment and may constitute infringement. The court said a trial remains warranted on the pirated copies issue and potential damages.

Reactions from authors, publishers, and Anthropic

Maria Pallante, CEO of the Association of American Publishers, told the court the judge “demonstrated a lack of understanding of how the publishing industry works” and called parts of the court’s timetable “troubling.”

Mary Rasenberger, CEO of The Authors Guild, who attended the hearing, pushed back on the judge’s suggestion that the guild might be pressuring authors, calling the insinuation “shocking” and saying its participation has been aimed at ensuring authors’ interests are fully represented.

Lead plaintiffs include thriller novelist Andrea Bartz and nonfiction writers Charles Graeber and Kirk Wallace Johnson. Johnson described the settlement as “the beginning of a fight on behalf of humans that don’t believe we have to sacrifice everything on the altar of AI.” Plaintiffs’ attorney Justin Nelson said the parties are confident the money will be fairly distributed and noted the case has received widespread media coverage. “This is not an under-the-radar warranty case,” Nelson said.

Anthropic said the settlement “will resolve the plaintiffs’ remaining legacy claims” and reaffirmed its commitment to developing safe AI systems. The company has also said it will destroy the original downloaded book files.

Commercial context – funding, valuation, and market ripple effects

Anthropic has recently secured a whoppingly large funding round of $13 billion in Series F that the company says values it at $183 billion, and has reported a rapid revenue run-rate increase into 2025. That investor backing and Anthropic’s commercial momentum are one reason the stakes in a trial or settlement are so large. The settlement could also influence ongoing suits against other AI companies, including OpenAI and Microsoft.

Separately, Microsoft plans to pay to use some Anthropic technology for certain Office 365 features, a sign of commercial traction that could shape how courts and markets view remedies and damages in AI-copyright cases. According to Reuters, Microsoft will pay its cloud rival Amazon Web Services to access the Anthropic models, according to the report. AWS is one of Anthropic’s largest shareholders.

What’s next?

Alsup ordered the parties to exchange a finalized works list by Sept. 15, submit a proposed claims form by Sept. 22, and return to court on Sept. 25 to determine whether preliminary approval is appropriate. If the judge remains unsatisfied, he warned that he may allow the December trial to proceed.

ⓘ As a part of our ongoing support for startups and SMEs, we publish feature and resource articles that may include external links to third-party websites. While such content is selected at our editorial discretion, LAFFAZ Media is not responsible for the content or practices of external websites. For more information, please refer to our Terms and Conditions.