Security weaknesses in software are the main technical problem in cybersecurity. Its timely identification is key to developing software capable of resisting cyberattacks. Static source code analysis (with large numbers of false positives) and dynamic application analysis (with limited coverage of possible execution paths) have been attempted for decades. This project consists of a type of static analysis that, instead of analyzing the source code, will analyze a textual description of the code generated automatically through AI techniques (e.g., https://denigma.app/). Basically, the idea is to explore to what extent such explanations are useful from a security point of view to detect weaknesses.