Hacker News
Anna's Archive Says AI Companies Are Buying Secondhand Books Through Intermediaries to Scan and Destroy Them for Pre-2022 Training Data
The post, at 359 points and 266 comments on its second submission, claims several AI companies are acquiring bulk secondhand books via intermediaries, scanning them destructively, and pulping the originals to secure text predating the 2022 generative-AI flood. Anthropic, Meta (81 terabytes of torrented ebooks) and Amazon are named, and the thread anchors on Judge William Alsup's ruling that Anthropic's buy-scan-destroy pipeline was transformative fair use provided no duplicate copies remained, alongside the $1.5 billion Anthropic paid to settle. Anna's Archive is soliciting donations to scan rare volumes first. Top commenters push back that most of this stock was already headed to recycling.
↳ Follow the thread