The goal of this paper is to show the importance of reliability for artificial intelligence (AI), which is in nowadays applying in almost every area of human life. Everything can fail, and AI is not an exception, as any other product or devices. AI also need to be reliable. The definitions of AI, reliability and some reasons for unreliability such as quality of data, small or insufficient number of training and test data, training on unreal data, nature of learning algorithms and user incompetence are analysed in this paper.