The invention discloses an application layer dynamic intrusion detection system and detection method based on artificial intelligence, wherein the detection system comprises an application layer gateway, a detection module, a judgment and operation module, a sample database and an updating module, the detection module comprises a detection model mixed with a convolutional neural network and a bidirectional long and short term memory neural network. The detection module after initialization is used for making an attack judgment on an application layer data packet, filtering the data packet above the threshold value and putting the data packet into a malicious sample database, and meanwhile, the data packet under the threshold value is not processed. The updating module is used for traininga new model by using the malicious samples and normal samples with a certain proportion in the sample database and updating the detection model in the detection module in real time. According to the invention, a universal detection method is used for the attack method of the application layer, the method has the advantages of high detection rate and low misjudgment rate. Meanwhile, the intrusion detection system has the advantage of dynamic updating model, and has good filtering effect on unknown zero-day attack.