Midjourney's secret list of 20,000 artists used to train AI models exposed in court

A leaked Google spreadsheet, filed as evidence in a class-action lawsuit against Midjourney, Stability AI, Runway AI, and DeviantArt, reveals nearly 20,000 artists whose copyrighted works were allegedly scraped to train generative AI models. The list, curated by Midjourney, includes 5,000 confirmed names and 15,000 suggested ones, spanning from Old Masters like Matisse and van Gogh to contemporary figures such as Anish Kapoor, David Hockney, Damien Hirst, and Banksy. The 24-page document, part of a 455-page amendment filed in November 2023, also contains Discord conversations where Midjourney CEO David Holz allegedly discussed circumventing copyright by using scraped datasets and conveniently forgetting their origin. The lawsuit, originally filed in January 2023 by illustrators Sarah Andersen, Kelly McKernan, and painter Karla Ortiz, had some claims dismissed by a California federal judge, prompting the submission of additional evidence. The spreadsheet, initially shared on social media and preserved via the Internet Archive, has since had its access restricted. The case highlights the ongoing legal battles over unauthorized use of artists' works for AI training.