The system relates to the field of computer software security. By computer software vulnerability detection, the invention discloses an automatic vulnerability detection method based on artificial intelligence directly on a binary language (namely a machine language), the method is named as an automatic artificial intelligence vulnerability detection system, and the automatic artificial intelligence vulnerability detection system comprises function identification based on artificial intelligence on the binary language, .On the premise that the function is identified in the binary language, the vulnerability is detected by applying an artificial intelligence-based method. According to the method provided by the embodiment of the invention, a user can perform vulnerability detection on purchased software on the premise that the user cannot take the source code, so that the problem of rapidly performing vulnerability detection on the closed-source software is solved. Meanwhile, whether the automatic artificial intelligence vulnerability detection system can train the model offline or not is detected online by utilizing the trained model, and particularly, rapid detection is carried out on large software, so that the problem that some existing software is very low in efficiency for the large software is avoided.