We test AI systems on taboo topics in AI safety & alignment, ensuring they hold up to our core principles.