Salesforce Accused of Using Pirated Books for AI Training
- Authors E. Molly Tanzer and Jennifer Gilmore have filed a lawsuit against Salesforce, alleging the company used pirated books to develop its XGen AI models.
- The lawsuit claims Salesforce initially acknowledged using the “RedPajama-Books” dataset in June 2023 but later removed references, describing data sources as “publicly available.”
- Salesforce CEO Marc Benioff has publicly stated that AI companies have stolen training data, reinforcing the lawsuit’s claims.
- The complaint alleges Salesforce used datasets like The Pile and Books3, which include over 196,000 books from Bibliotik.
- Legal expert Ishita Sharma noted authors need to prove financial harm beyond just usage of their works for training.
Salesforce faces legal challenges over alleged copyright infringement in developing its XGen AI models using datasets containing pirated books. The case highlights ongoing concerns about the legality of data used in training AI models and the necessity for authors to demonstrate financial harm.
This lawsuit underscores tensions between content creators and tech companies over intellectual property rights in AI development, with Salesforce accused of exploiting copyrighted materials without permission or compensation. (Source)
