The case is one of dozens brought by copyright owners including authors and news outlets against tech companies over the training of their large language models.