Lawsuit alleges Zuckerberg OK’d training AI on copyrighted works
Meta CEO Mark Zuckerberg approved the use of a dataset containing copyrighted material to train the company’s Llama AI model, a new filing in a copyright suit against the company says.
A last-minute discovery filing revealed that Meta engineers had reservations about the legality of using the LibGen dataset, acknowledging that it was “a dataset we know to be pirated.”
Engineers also wrote scripts to remove information related to copyright from the dataset before training, according to the court filing.
Lawyers for the authors who filed the suit against Meta said that the question of using LibGen was escalated to “MZ.” The team was then given the go-ahead to use the copyrighted material.
Engineers also wrote scripts to remove information related to copyright from the dataset before training, according to the court filing.
Lawyers for the authors who filed the suit against Meta said that the question of using LibGen was escalated to “MZ.” The team was then given the go-ahead to use the copyrighted material.