Anthropic to pay $1.5B to authors in largest copyright settlement in US history

A historic settlement reshapes AI and publishing

Anthropic, the fast-growing AI startup behind the Claude chatbot, has agreed to pay at least $1.5 billion to settle a lawsuit brought by authors.

The case alleged that Anthropic trained its models using pirated books downloaded from sites like Library Genesis and Pirate Library Mirror. This will be the most significant copyright settlement in U.S. history if approved.

The agreement marks a turning point in the clash between creative industries and AI companies that depend heavily on written works to train large language models.

Authors allege mass-scale copyright infringement

The lawsuit accused Anthropic of taking millions of digitized books from piracy sites without permission and feeding them into Claude’s training data.

Roughly 500,000 works are covered by the settlement. Each eligible author will receive about $3,000 per book, a much higher rate than previous copyright settlements. The plaintiffs argued that this “systematic theft” devalued their creative labor while giving Anthropic a competitive edge.

The case quickly became symbolic of broader concerns about how AI startups acquire and process training data.

The authors behind the lawsuit

The lawsuit was initially filed by three writers: Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson. Bartz is a novelist known for thrillers like The Lost Night, while Graeber and Johnson are award-winning nonfiction authors.

Their case later expanded to represent hundreds of thousands of published works.

By targeting Anthropic, one of the highest-valued AI startups, the authors aimed to send a message to the entire industry: creative labor must be respected, even in the race to build transformative technologies.

The settlement amount is unprecedented

At $1.5 billion, this is the largest publicly reported copyright settlement ever in the U.S., surpassing any class action or individual recovery in history.

Legal analyst Justin Nelson, who represented the authors, called the deal “nothing short of remarkable.” He noted that the per-work payout of $3,000 far exceeds expectations, especially after duplicates and non-copyrighted works were excluded.

For many authors, this represents the first meaningful financial acknowledgment that their work has real value in the era of generative AI.

A mixed ruling set the stage

Earlier in the case, Judge Alsup granted summary judgment for Anthropic on training with lawfully obtained books, but allowed claims related to pirated copies to proceed.

With a trial scheduled for December, potential damages were expected to be substantial, though estimates varied. Doing so avoided a courtroom fight that could have been financially devastating.

Payments will be structured in stages

Court documents show that Anthropic will make four payments into the settlement fund. The first, $300 million, is due within five business days of final court approval.

If the list of infringed works exceeds 500,000 titles, Anthropic will owe an additional $3,000 per extra book. This sliding scale ensures that the settlement amount grows with the actual scope of infringement.

For authors, it provides confidence that compensation won’t be capped if more of their works are found in Anthropic’s datasets.

Anthropic agrees to destroy pirated files

As part of the settlement, Anthropic must permanently delete the pirated book files that were initially downloaded.

Court filings revealed the company had stored over seven million pirated works in a central library, including about five million from Library Genesis and about two million from Pirate Library Mirror.

While Anthropic has since shifted to licensing and scanning books legally, the requirement to destroy past datasets underscores a new industry standard. Companies cannot simply keep using unlawfully sourced data while arguing fair use.

The role of Books3 in training data

One revealing detail was the inclusion of the Books3 dataset, a collection of around 200,000 digitized books assembled by independent researchers, among whose works was Andrea Bartz’s The Lost Night.

This highlighted how easily individual authors’ creations could be swept up into massive datasets without consent or payment, fueling outrage and strengthening the case against Anthropic.

Authors Guild praises the outcome

Mary Rasenberger, CEO of the Authors Guild, called the settlement “an excellent result for authors, publishers, and rightsholders.”

She said it sends a strong message that pirating creative works for AI training has consequences. For years, the Guild has warned that writers were being exploited as free raw material.

The $1.5 billion figure validates those concerns and shows that authors can prevail even against multibillion-dollar AI companies. It also sets a powerful precedent for future cases against rivals like OpenAI.

European authors feel left out

Not everyone benefits equally from the deal. Some groups in Europe have pointed out that the settlement is limited to U.S. class members (or works covered by the U.S. lawsuit), meaning many

European authors whose works were included in Anthropic’s datasets may be ineligible for payments. It confirms piracy by AI companies comes at a cost, but it offers no direct relief. This underscores the need for international copyright frameworks in the AI age.

Critics see settlement as a cost of business

Some industry watchers argue that Anthropic may view the huge settlement as the cost of doing business. The company recently raised $13 billion in new funding at a $183 billion valuation.

While Anthropic’s valuation and funding have attracted considerable attention, clear figures regarding its revenue expectations for this year have not been confirmed in reliable public filings.

The broader publishing industry watches closely

The settlement has significant implications for ongoing lawsuits. High-profile authors like George R.R. Martin, John Grisham, and Jodi Picoult have sued OpenAI, alleging “systematic theft on a mass scale.”

Newspapers, including The New York Times, have also filed copyright claims. The Anthropic deal shows plaintiffs can secure massive payouts without going to trial.

That will likely embolden other creators to keep pushing. For the publishing world, it could spark more collective action to hold AI firms accountable.

Fair use remains a contested idea

Anthropic argued throughout the case that its training fell under fair use, since AI systems transform text rather than reproduce it verbatim.

Judge Alsup partially agreed, calling the process “quintessentially transformative.” However, he drew a hard line on how data is acquired. Simply scraping or downloading pirated books is not permissible.

This distinction may shape future copyright law: courts could allow AI training on licensed or public-domain works but penalize companies that cut corners using pirated data.

Internal doubts surfaced at Anthropic

Court documents revealed that even Anthropic employees worried about the legality of relying on pirated books. Internal discussions flagged risks, prompting the company to hire Tom Turvey, a former Google Books executive, to build a more sustainable data pipeline.

Under his guidance, Anthropic began legally purchasing books, scanning them, and incorporating them into its datasets.

The pivot showed that AI firms can train responsibly, but it also underscored that early shortcuts exposed companies to enormous liability.

Settlement prevents a possible corporate crisis

Had Anthropic gone to trial and lost, the damages could have been devastating, possibly in the tens of billions. Some legal analysts suggested the company risked collapse if the court found willful infringement across millions of works.

By settling, Anthropic avoided a potentially existential threat. The deal allows it to preserve investor confidence, maintain market momentum, and compete with rivals like OpenAI and Google, the tradeoff: paying dearly now while trying to rebuild trust with authors and publishers.

See how Anthropic is pushing forward with Claude AI by letting users build apps directly on the platform.

The future of AI depends on fair partnerships

Ultimately, this case shows that AI and creativity must coexist. Authors don’t want to stop innovation but demand respect and compensation.

If AI companies embrace licensing, collaboration, and fair agreements, they can access the data they need without undermining creators.

Anthropic’s settlement may sting financially, but it also offers a roadmap: the path forward lies not in scraping pirated works, but in building sustainable partnerships that allow AI technology and human creativity to thrive.

Find out why Apple is weighing partnerships with OpenAI or Anthropic to power the next generation of Siri.

What do you think about the Anthropic lawsuit for violating the terms and conditions by exposing user data? Please share your thoughts and drop a comment.

Read More From This Brand:

• AI design just got real with Anthropic’s Claude, a new Canva superpower
• Google, OpenAI, and Anthropic researchers say AI models are becoming too complex
• xAI faces backlash as OpenAI and Anthropic call out ‘reckless’ practices

Don’t forget to follow us for more exclusive content on MSN.

If you liked this story, you’ll LOVE our FREE emails. Join today and be the first to get stories like this one.

This slideshow was made with AI assistance and human editing.