Midjourney May Withhold Some Datasets Used to Train AI (1)

Artists whose work was allegedly used by Midjourney Inc. to train its generative AI product aren’t entitled to access all datasets used in the training process during discovery in their copyright infringement suit, a federal court said.

Only the datasets Midjourney received from Large-Scale Artificial Intelligence Open Network are relevant to the artists’ claims, US Magistrate Judge Lisa J. Cisneros said June 27 for the US District Court for the Northern District of California. Making Midjourney produce more would be disproportionate to the needs of the case and be contrary to the limits created by the Federal Rules of Civil ...