A class-action lawsuit against AI company Anthropic over copyright infringement is nearing settlement, with both parties reaching an agreement in principle[1]. The lawsuit, filed by authors Andrea Bartz, Kirk Wallace Johnson, and Charles Graeber, alleged Anthropic illegally downloaded millions of books to train its AI models[2].
U.S. District Judge William Alsup certified what could be the largest copyright class action ever, potentially including up to 7 million claimants[1:1]. The lawsuit claimed Anthropic pirated books from online sources including Books3, Library Genesis, and Pirate Library Mirror[2:1].
“This historic settlement will benefit all class members,” said Justin A. Nelson, attorney for the authors[1:2]. The parties must file a motion for preliminary approval by September 5, 2025[1:3].
While settlement terms remain undisclosed, the case had serious implications - industry advocates warned that if every eligible author filed a claim, it could “financially ruin” the AI industry[1:4]. Anthropic had previously argued the lawsuit threatened its survival as a company[1:5].
Fuck copyright. I don’t think individuals should have to pay - even with their private data - to access information, and that means companies shouldn’t either. Especially ones providing a public service to people with open weights AI models.
I don’t think individuals should have to pay - even with their private data
Agree.
[…] and that means companies shouldn’t either.
Disagree.
Whn a person pirates, they usually do it for a) themselves, b) their family or c) a close friend. Some might share on a larger basis.
And other than that, they also usually use it for a) educational or b) entertainment purposes.
For companies, it’s alsmost always d) On a larger basis and c) commercially.
As most licences and contracts differentiate the two uses, so should the law.
The fact that I can download a book online and read it (sneakily, and technically illegally) doesn’t mean that if I became an AI LLC I could download it, along with thousands of others, to then sell as my AI’s “knowledge”.
Making that an AI’s knowledge is “storing in a retrieval system” and commercial use isn’t a free use criterion.
The true problem with (common law) copyright is the fact that it can be bought and sold. Or rather, the author doesn’t own it - the publisher does. Which goes against the initial idea of the author getting dividends from their works.
In a world where people require money to survive, copyright protections ensure writing and other arts remain a realistic (if rarely profitable) activity for those without the luxury of complete financial independence and a wealth of spare time.
