Methods to ensure AI systems follow human values and goals. Important for safe and reliable behavior.