Meta tries to scapegoat staff who downloaded torrents to spare execs

Meta is defending its AI training practices in court by arguing that the unauthorized use of pirated books for its models was the result of individual employee actions rather than a deliberate corporate policy. The company maintains that staffers downloaded torrented datasets without management direction, positioning these incidents as isolated errors to distance leadership from copyright infringement allegations. This legal argument attempts to shield the firm from broader liability regarding how it sourced data to train its Llama models.