What it Means that a Judge Approved Anthropic's $1.5 Billion Settlement with Authors — July 21, 2026

Radar Flash Edition. Federal Judge Araceli Martínez-Olguín finally approved yesterday, Monday, July 20, the settlement in which Anthropic will pay $1.5 billion to the authors and publishers who sued the company for using their books to train Claude. It is the largest known copyright settlement in U.S. history: approximately $3,000 per work for nearly 500,000 works, plus $101 million in fees for the plaintiffs' attorneys—significantly less than the $187.5 million they requested. The judge rejected objections from authors who considered the figure insufficient as they were not "based on a realistic assessment of the risks and rewards of a trial," and according to Anthropic, more than 91% of covered authors and publishers have already claimed their share. Source

Why it matters: the legal nuance is more interesting than the figure. Judge William Alsup had already ruled in June 2025 that training an AI with legally acquired books constitutes "fair use"; what sank Anthropic was having downloaded and stored more than seven million pirated books to build its library. By settling instead of going to trial, there will be no appeal or binding precedent: each judge remains free to decide in the open cases against OpenAI, Meta, Google, and Midjourney—including the recent class-action lawsuit by publishers and authors against Google regarding Gemini training. The bill for training data has just become a real line item in the cost of building AI, and this will eventually be reflected in industry pricing.

For your company: three practical takeaways. First: using Claude, ChatGPT, or any other generative AI does not expose you to anything due to this case—the lawsuit was about how Anthropic obtained the data, not how customers use it. Second: even so, if you contract AI on a business plan, check that the contract includes intellectual property indemnification (Anthropic, OpenAI, Microsoft, and Google offer this in their commercial plans)—it is the clause that covers you if a third party claims against the outputs you generate. Third: if you train or fine-tune models with your own or third-party data, document the source and licenses of each origin—the penalty was not for training, it was for pirating the source material, and that distinction applies equally to an SME fine-tuning a model with PDFs "found" on the internet. Today's edition of the Daily Radar has the rest of the news.

Frequently Asked Questions

What exactly did the judge approve in the Anthropic case?

Federal Judge Araceli Martínez-Olguín (Northern District of California) gave final approval on July 20, 2026, to the $1.5 billion settlement between Anthropic and the authors who sued for copyright infringement: approximately $3,000 per work for nearly 500,000 works, the largest known copyright settlement in the US. The deal was signed in September 2025 and now, with objections rejected, payments can be executed.

Does this affect companies using Claude or other generative AI?

Not directly: the case was about how Anthropic obtained the books (pirated downloads), not the use of the product, and customers are not parties to the suit. The practical recommendation is contractual: in business plans, check that your AI provider offers intellectual property indemnification, and if you train models with your own data, document the licenses for each source.

Does this set a precedent for other AI and copyright lawsuits?

It is not binding. By settling, Anthropic avoided a trial and appeal, so there is no higher court doctrine: the preliminary ruling remains that training with legally acquired material may be fair use, but each judge will decide in the open cases against OpenAI, Meta, Google, or Midjourney. What it does set is a reference price: training data is no longer free.