Bloomsbury Publishing secures millions from Anthropic settlement over AI training on copyrighted works

Bloomsbury Publishing Among Major Beneficiaries of Historic Anthropic Settlement

Bloomsbury Publishing confirmed it will collect a substantial share of the $1.5 billion Anthropic settlement, marking the largest copyright payout in US history 1

1

. The London-based publisher, best known for the Harry Potter series, disclosed that 14,087 of its titles are covered by the deal, with each work drawing compensation of approximately $3,000 before deductions for legal fees and court costs 1

1

. After a deduction of about 10% for attorney fees and other expenses, Bloomsbury Publishing stands to receive roughly $21 million, with proceeds split equally between the company and affected authors 5

5

.

US District Judge Araceli Martínez-Olguín granted final approval Monday to the class-action lawsuit settlement in Bartz v. Anthropic, a case that stemmed from the AI company's unauthorized use of copyrighted books to train its Claude AI models 3

3

. The three named plaintiffs—novelist Andrea Bartz and non-fiction writers Charles Graeber and Kirk Wallace Johnson—first struck the agreement with Anthropic in September 2025 . About 91% of the 482,000 works covered in the suit have been claimed, according to the authors' lead lawyer Justin Nelson, who described it as "the largest known copyright recovery in history" 2

2

.

Copyright Infringement Claims Center on Pirated Book Repositories

The settlement resolves one of the first significant legal reckonings over how generative AI systems are built. While Anthropic had bought and scanned physical books legally, the company also pulled roughly seven million copies from pirate repositories—Library Genesis in 2021 and the Pirate Library Mirror in 2022 1

1

. In June 2025, Judge William Alsup ruled that AI training on copyrighted text could count as fair use, while acquiring pirated copies could not 1

1

. This distinction did most of the work in shaping the case's outcome.

Anthropic chose to settle rather than appeal, which leaves the fair-use finding as a single district-court opinion rather than binding precedent 1

1

. The company did not admit wrongdoing in agreeing to pay. The arithmetic favored settlement, since statutory damages can reach $150,000 per infringed work—an exposure that, spread across a class this size, dwarfed the $1.5 billion on the table and made a jury trial the larger gamble 1

1

. After duplicates and ineligible works were stripped out, about 500,000 distinct titles qualified for the class, each needing a US copyright registration to count 1

1

.

Stock Price Rises as Market Responds to Settlement News

Bloomsbury Publishing stock climbed 1.4% to trade at 651 pence following the court's final approval, providing investors with a concrete reason to buy the London-listed publisher 3

3

. The company expects to receive the cash in instalments beginning in the second half of its financial year, according to its regulatory filing 1

1

. The settlement news arrived against a backdrop of already-positive company momentum, with Bloomsbury recently backing its full-year guidance and citing benefits from ongoing AI licensing revenue and growth in its Academic & Professional division 3

3

.

The company noted it had net cash of £29.2 million as of its February 28 financial year-end 5

5

. Bloomsbury struck an AI partnership with Google in September of last year and has benefited from AI licensing revenue in the early months of its new fiscal year, underscoring how AI monetization of its content catalogue has become a meaningful and recurring driver of its financial performance 3

3

. The consensus analyst rating on Bloomsbury remains "Strong Buy," with all five covering analysts recommending the stock 3

3

.

Broader Implications for Publishers and Authors in AI Era

For publishers, the Anthropic AI settlement payout lands as they press near-identical claims elsewhere. Five houses are suing Meta over Llama, and a separate group has taken Google to court over Gemini, each arguing their catalogues were ingested without permission or payment 1

1

. Those cases remain free to reach conclusions of their own, since the Anthropic ruling binds no one 1

1

. Copyright has become a major battleground in the AI boom because tools like Anthropic's suite of Claude chatbots are trained on vast amounts of data taken from the open web, including copyright-protected work such as novels and news articles 2

2

.

US-based AI companies have argued that using such work is allowed under the legal doctrine of fair use, which allows use of copyrighted work without the owner's permission in certain circumstances 2

2

. However, creative professionals and artists around the world have been demanding that AI companies seek permission first—or at least pay them for using their work 2

2

. When the payments clear, they will be among the first tangible transfers of money from an AI company to the writers and publishers whose books trained large language models 1

1

. Whether this arrangement becomes a template or stays a one-off now rests with courtrooms that have yet to rule on similar cases involving other AI models and tech giants.