Pascal's Chatbot Q&As
Posts
Judge William Alsup of the Northern District of California granted preliminary approval to a $1.5 billion settlement in Bartz v. Anthropic—a case that has already earned its place in copyright history

Judge William Alsup of the Northern District of California granted preliminary approval to a $1.5 billion settlement in Bartz v. Anthropic—a case that has already earned its place in copyright history

It's the largest recovery in a copyright dispute, dwarfing earlier settlements involving music, film, and publishing sectors. Building the future won’t excuse borrowing the past without permission.

Pascal Hetzscholdt
September 25, 2025

The Anthropic Authors Settlement: A Historic Moment in Copyright and AI

by ChatGPT-4o

On September 25, 2025, Judge William Alsup of the Northern District of California granted preliminary approval to a $1.5 billion settlement in Bartz v. Anthropic—a landmark case that has already earned its place in copyright history. It is, by all accounts, the largest recovery ever in a copyright dispute, dwarfing earlier settlements involving the music, film, and publishing industries.

What the Case Was About

The plaintiffs—authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson—filed a class action in 2024 alleging that Anthropic, the AI startup behind the Claude assistant, had trained its large language models using books obtained from pirate sites like Library Genesis and Pirate Library Mirror.

The claims struck at the heart of AI’s data hunger: whether companies can freely use copyrighted works without permission, so long as they’re “only” for training. Authors and publishers argued this was textbook infringement.

The Settlement Terms

Anthropic’s proposed settlement, now preliminarily approved, includes:

$1.5 billion payout – roughly $3,000 per book plus interest, distributed among authors who make claims.
Dataset destruction – Anthropic agreed to delete training sets that contained pirated books.
Notice and claims process – a structured timetable where authors can register, opt out, or file claims (see below).

Anthropic, valued at $183 billion, framed the deal as a way to “resolve narrow claims” and refocus on safe AI development. Authors, for their part, called it a step toward “real accountability” and a warning shot to every other AI firm: creators’ rights cannot be bypassed.

Judge Alsup’s Role

Judge Alsup, known for his rigorous oversight in complex tech cases (he presided over Oracle v. Google), initially held back approval. He had raised more than 30 detailed questions about fairness, notice, and distribution. Only after the plaintiffs’ lawyers addressed his concerns did he bless the deal as “fair, reasonable, and adequate” under Rule 23(e)(2).

He also noted he may go inactive by year’s end, meaning final approval could fall to another judge.

Timeline of Next Steps

The settlement process now unfolds over the next several months.

Notice period begins (October 2025): Settlement administrator launches publication, digital, and social media notices.
Publisher cooperation (by November 2025): Major publishers must supply author contact information so notices reach class members.
Direct notice complete (November 2025): Authors receive official communications about eligibility.
Opt-out deadline (January 2026): Authors who don’t want to participate must formally opt out.
Claim submission deadline (March 2026): Eligible authors file for their share of the settlement.
Final approval hearing (April 2026): Court reviews the full process and considers any objections.
Distribution (June 2026): Assuming no appeals, the administrator calculates per-work and per-author amounts, and checks are cut.

Why This Matters Beyond Anthropic

This isn’t just about one company—it sets a benchmark for every ongoing AI copyright lawsuit. Cases against OpenAI, Meta, and others now have a reference point for damages and settlement structure.

The deal also creates a de facto price tag for books in AI training: ~$3,000 per title. While far from a negotiated license, that number will ripple through future bargaining between AI firms and rights holders.

Moreover, the dataset destruction requirement signals that courts may not accept a “pay once, keep training forever” approach. If upheld, it pressures AI companies to secure legitimate data pipelines going forward.

What to Watch

Appeals risk: Objectors or opt-outs could challenge the fairness of the distribution, potentially delaying payment.
Other sectors: Musicians, journalists, and visual artists will study this case for strategies and valuation cues.
Regulatory echoes: Policymakers in the U.S. and abroad may use this as a reference point for AI copyright legislation.

Conclusion

The $1.5 billion Anthropic settlement is more than a payout—it’s a signal flare in the global struggle to align AI development with copyright law. For authors, it delivers recognition and compensation. For Anthropic, it’s costly but clarifying. And for the AI industry at large, it’s a sharp reminder: building the future won’t excuse borrowing the past without permission.

The next months will reveal whether this agreement becomes a durable template or just the first of many expensive lessons in AI’s copyright reckoning.