As Benioff faces backlash in SF, Salesforce hit with copyright suit for AI training allegedly on pirated books. Suits hit 53 in US.

Marc Benioff sparked backlash this past week for his comments in favor of bringing in the National Guard to San Francisco. Benioff later did damage control on Twitter. But it might be too little too late.

Marc Benioff faces backlash after suggesting National Guard deployment to SF

His remarks have sparked a wave of backlash from Bay Area officials and residents, days ahead of the company’s major annual conference, Dreamforce.
In a phone interview with The New York Times from his… pic.twitter.com/VpJ0DpTptU
— Betty Yu (@bett_yu) October 12, 2025

Ron Conway quits Salesforce Foundation board after Marc Benioff's National Guard comments:
"It saddens me immensely to say that with your recent comments, and failure to understand their impact, I now barely recognize the person I have so long admired.” https://t.co/7utY9NVJT0
— Biz Carson (@bizcarson) October 16, 2025

Comedians Kumail Nanjiani and Ilana Glazer abruptly withdrew from Dreamforce as Marc Benioff faces backlash for calling for the National Guard to be deployed in San Francisco. https://t.co/dbVIroBJ8w
— San Francisco Chronicle (@sfchronicle) October 16, 2025

@nbcbayarea
Salesforce founder Marc Benioff appears to be walking back comments he made last week in which he welcomed the deployment of National Guard troops to San Francisco.#bayarea #sanfrancisco #military #trump
♬ original sound – NBC Bay Area – NBC Bay Area

To make matters worse, yesterday, book authors E. Molly Tanzer and Jennifer Gilmore, represented by the Joseph Saveri Law Firm, based in San Francisco, filed a class action copyright infringement lawsuit against Salesforce. The Complaint alleges that Salesforce used pirated books datasets to train its AI model(s).

33. In March 2022, Salesforce released its CodeGen series of LLMs.
“[a]ll variants of CODEGEN are firstly pre-trained on the Pile.”7 Within The Pile is the Books3 dataset containing Plaintiffs and Class members’ works. Accordingly, Salesforce downloaded and stored a copy of The Pile containing Books3 before March 2022 and used it to develop and train its CodeGen series of LLMs.

34. With this library of nearly 200,000 pirated books at its disposal, Salesforce decided to use Books3 for the development and training and of its next language model.

35. In June 2023, Salesforce released the XGen series of LLMs in a blog titled, “Long Sequence Modeling with XGen: A 7B LLM Trained on 8K Input Sequence Length” (“XGen Blog”). series consists of the XGen-7B-4K-base, XGen-7B-8K-base, and XGen-7B-{4K, 8K}-inst models. In its announcement, Salesforce provided a chart detailing the models’ pre-training data, which admits that “RedPajama-Books” was one of the training datasets:

…

49. For example, in a January 2024 interview, Salesforce CEO Marc Benioff acknowledged that artificial intelligence companies “ripped off” training data to build their models, and that “all the training data has been stolen.”20 Benioff also stated, “There’s a pretty great company to be built on a standardized set of training data that lets all these companies play a fair . . . game and let the content creators . . . get paid fairly for their work. And I think that bridge has not yet been crossed. And that’s a mistake by the AI companies. Very easy to do.”21

50. Benioff is right—technology companies like Benioff’s own Salesforce that use the intellectual property of copyright holders like Plaintiffs and Class members should fairly compensate them. And that is why Plaintiffs bring this lawsuit.
complaint in Tanzer v. salesforce