What's this about?
A federal judge gave final approval to Anthropic's $1.5 billion settlement with authors whose books were downloaded from pirate libraries and used in training data. Eligible authors/publishers get about $3,000 per title, and the judge cut the class lawyers' fee in half.
What exactly was illegal here?
Not the training. Courts have leaned toward AI training on lawfully obtained books being "fair use" — the violation was *how the books were obtained*: mass-downloading from pirate sites instead of buying or licensing copies. Think of the difference between photographing books you bought versus shoplifting them first.
What's the community saying?
HN 191 points, 145 comments
- "$1.5B is a rounding error" — many argue the fine is just a cost of doing business and won't deter anyone.
- Clarification upvoted repeatedly: the ruling is about piracy, not about whether training itself is legal.
- Authors' math: $3,000 per title, typically split 50/50 with the publisher under standard contracts.
- Open question: do open-weights labs trained on the same pirated datasets face the same exposure?