The Models Are Yours: The Public's Leverage in AI

The author argues that if courts find AI training on copyrighted works infringes copyright, the public and creators gain major leverage over AI development. The argument rests on the fair use factors, with the author contending that trained models are lossy compressed copies of their training texts, comparable to MP3 files, and that they harm the market for original works. The piece also cites statutory damages of at least $750 per violation as a source of potential liability, and it proposes coalition-building and a new copyright framework as ways for authors to organize.