In the previous chapter, AI for cyberspace security was introduced in detail. As a new technology, we seek to find out whether AI itself exist intrinsic security problems? From the internal perspective, an AI system is composed more complicated components, such as data, framework, algorithm, model, and operation. It is obvious that any vulnerable parts would cause security problems in such a complicated system, and we call such vulnerability in AI systems as AI intrinsic security (AI security for short). In this chapter, we introduce AI security from the following aspects, data security, framework security, algorithm security, model security and operational security.