Report: Claude trained on pirated books by Stanford professors
AI models learn from huge piles of text, and the authors often never said yes. On Monday, we saw Elvis Presley's rights holders sue Anthropic over song lyrics. Today: same company, new case.
The Stanford Daily reports that Anthropic trained Claude on illegally obtained books written by Stanford professors. The story got 523 points and 274 comments on Hacker News. Another page also climbed Hacker News: Claude's own help article 'Is my data used for model training?', with 219 points. People clearly want to know where their words end up.
The timing is hard for Anthropic, just as it prepares to sell shares to the public. More authors will likely check if their books were in the pile.
The company that sells itself as the careful lab still faces questions about how it got its training data. If you write, teach or create, your work may already be inside these models.
