AI Guarded Code Editor: Verifying Authentic Human Input

AI Tools such as ChatGPT, GPT-4 and GitHub Copilot change the very nature of how developers build software by greatly helping in the quick generation of code. While this proves to be efficient for professionals, it poses challenges in the competitive programming space. Generative AI is greatly misused by candidates in coding tests and online interviews. We will design a guarded code editor that detects the probability of usage of GenAI in code submissions with specialized AI detection models that use unique metrics, such as long AI-generated keywords, unusual number of utility functions, and very verbose comments that have been shown to be associated with usual AI-generated content. We will focus on stylistic and structural markers such as redundant comments, too general or too redundant variable names, and predictable formatting to realize a lightweight and scalable detection framework that fits into a user-friendly code editor. Our system also includes mechanisms for immediate feedback that inform the users on the reasons for differences in AI-generated and human-written code, thereby promoting better integrity of code in their own practices. This solution addresses a few critical concerns identified in some recent work that detection tools have limitations to the availability of responding to a domain-specific method responsive to how it would likely be combated using misapplication of the AI system. In other words, with a code editor that does not accept the direct submission of AI-generated code, the proposed solution will most definitely prove effective in both bringing forward practicality and ethical means while ensuring originality and improvement of quality in academic and competitive programming.