Anthropic’s landmark $1.5B copyright settlement is approved
TechCrunch Kirsten Korosec ● Covered by 3 sources
A federal judge just finalized Anthropic's $1.5B settlement with authors and publishers over pirated books. It's the biggest copyright payout ever, but the industry's core legal fight is still wide open.
Based on reporting by TechCrunch, Kirsten Korosec — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic can finally start writing checks. Judge Araceli Martinez-Olguin signed off Monday on the AI company's $1.5 billion settlement with a class of authors and publishers, closing out a case that began under her predecessor, Judge William Alsup, who retired after granting preliminary approval last year. The math is simple and staggering: $3,000 per work across roughly 500,000 works, making it the largest copyright settlement in U.S. history, according to Reuters.
The money is real, but the win for authors is murkier than the headline number suggests. Alsup's original ruling actually sided with Anthropic on the question that matters most going forward: training an AI model on copyrighted text, he said, counts as fair use. That's a significant marker for the entire industry, one that Anthropic will happily cite elsewhere.
What got Anthropic in trouble wasn't the training itself, but the sourcing. The company built its book library two ways, one legitimate and one not. It bought and scanned some books, which Alsup had no problem with. But it also pulled titles from pirate sites like Library Genesis and Pirate Library Mirror, and Alsup ruled that piracy claim serious enough to send to trial. Facing a jury and open-ended damages, Anthropic chose to settle instead.
Because it settled rather than appeal, the fair-use ruling stays confined to this one case and this one judge. It's not binding on anyone else, and other courts are already treating similar questions differently. Lawsuits over AI training data are still working through the system against Google, Meta, Midjourney, and OpenAI. Just last week, a group that includes Hachette, Cengage, Elsevier, Scott Turow, and S.C.R.I.B.E. sued Google, arguing it trained Gemini on their copyrighted work without permission — proof that this fight is nowhere close to settled industry-wide, pun very much intended.
My take — AI-written commentary, not fact-checked reporting
A $1.5 billion check sounds like justice until you notice what it actually punished: piracy, not AI training itself. Courts keep signaling that scraping copyrighted text to train a model is fine as long as you didn't steal the copy first, which is a pretty convenient line for companies with big enough budgets to just buy legitimate access later. Authors got paid, but the precedent they actually wanted — that training itself requires permission — never showed up, and it won't, since Anthropic made sure this case never reached an appeals court. Expect every other AI company staring down a lawsuit to take the same lesson: settle fast, keep the fair-use ruling intact, and treat the payout as a cost of doing business rather than an admission of anything.”}]}
Read more about this at: TechCrunch
Related stories
“Zlibrary my beloved”: Anthropic staff chats extolling piracy cited in Sony suit
Ars Technica · 2 weeks ago ·
8