Sony Music Publishing and Warner Chappell filed suit against Anthropic, Dario Amodei, and co-founder Benjamin Mann on August 28. What's unusual is that the underlying facts aren't in dispute anymore.
Last September, in the Bartz case, a federal judge ruled that training an AI model on copyrighted text was legal, but downloading the training copies via piracy was not. Anthropic settled that case for $1.5 billion after admitting Mann personally torrented over five million books from Library Genesis in 2021, and staff pulled two million more from Pirate Library Mirror in 2022.
Sony and Warner's complaint cites those exact same downloads, now tied to MusixMatch and LyricFind lyric datasets. They're not asking a court to rule on anything new, they're applying a ruling that already exists to a different set of copyrighted works. Statutory damages run $150,000 per work, so the number could dwarf the book settlement depending on how many songs are in scope.
What I don't have a good answer for: once a company settles one IP class action over a specific data-acquisition method, does that admission become effectively permanent exposure for every other rightsholder whose work touched the same pirated corpus? Is there a legal mechanism that closes that door, or is Anthropic just going to keep getting sued by whoever's catalog turns up in the same torrent logs?
[link] [comments]