Revealing Court Documents: Meta’s Risky AI Training Tactics Using Copyrighted Content