• by Ray Schultz , 73 minutes ago

A $10 billion+ lawsuit has been filed by nine newspapers, accusing OpenAI and Microsoft of violating their copyrights for the purpose of training large language models. 

The case, one in a long line of such actions, was filed Wednesday by California Newspapers Partnership, Prairie Mountain Publishing Company LLP; MNG-BH Acquisition LLC; Hartford Courant Company, LLC; The Daily Press LLC; The Morning Call, LLC; Virginia-Pilot Media, LLC; Los Angeles Daily News Publishing Companies; and the San Diego Union-Tribune, LLC. 

The complaint argues that, “unlike a traditional search result, the synthetic output does not include a prominent hyperlink that sends users to the Publishers’ website. Rather, the output disguises the results as the work of the GPT system itself.”

It continues, “There is no question that the Defendants’ models have ‘memorized’ the pilfered copies of the authors’ and publishers’ copyrighted works. And in order to remain current, the Defendants cannot rely just on the content they stole in the past – they have to update their models regularly with new material so they can provide their users with the latest information.”

The complaint adds, “Hundreds of thousands of the Publishers’ Works were copied and ingested—multiple times—for the purpose of “training” Defendants’ GPT models. 

In addition, the complaint notes that “the publishers of nine regional newspapers join the long list of publishers and authors who have filed lawsuits against OpenAI, Microsoft, and other AI companies. Most of these lawsuits have been consolidated in this Court, and have survived motions to dismiss largely intact.” 

The case is on file with the U.S. District Court for the Southern District of New York.