The invention provides a method and a system for hierarchical detection of software behavior defects based on an artificial intelligence decision tree, First, test the software, After testing, A decision tree of artificial intelligence is constructed by using the historical data, Extracting classification rules of software behavior, Then, during the running of the software, Aiming at the softwarebehavior during software running, a layered detection model of software behavior defects based on system layer, state layer and function layer is established for the information system. The risk behavior and abnormal behavior of the system layer and the state layer are detected first, and when there is no software behavior defect in the system layer and the state layer, the function layer detection is carried out again. By comparing the method of the invention with the traditional defect detection method, the time for accurately locating the cause of the defect after the defect is found in thesoftware running process is obviously longer than that of the traditional detection method.