Red Teaming Language Models to Reduce Harms • Libertify