Authors Claim Salesforce Used Pirated Books to Train AI Models

Two fiction authors are suing Salesforce Inc. alleging the company unlawfully downloaded, stored, and copied books to develop its artificial intelligence language models.

Salesforce’s training datasets contain Book3 corpus, a data set holding “hundreds of thousands of copyrighted books that were acquired without the authorization or consent of the authors,” E. Molly Tanzer and Jennifer Gilmore said in a Wednesday complaint filed in the US District Court for the Northern District of California.

A spokesperson for Salesforce said the company doesn’t “comment ...