“Meta has spent the last five to seven years developing sophisticated AI systems to identify and classify content across 18 different categories of harm.”