Large Language Models Are Built Without Guilt