The Impact of Security Mindset on the Use of AI Assistants in Computing Education

The recent advances in large-language models (LLMs) have started shifting the way computing-related students write codes. LLM-based AI assistants, such as ChatGPT and Copilot, are now increasingly adopted to produce functional code by computing-related students. Although studies have shown that these AI assistants can improve coding efficiency, they also raise security challenges, especially when users lack a security mindset. Given the fact that AI assistants are increasingly integrated into computing education, this article performed an empirical study to explore the impact of a security mindset on the use of AI assistants for computing-related students in their coding and development. Our three-stage study showed that a significant portion of computing-related students currently lack security awareness toward the use of AI assistants. In addition, their usage of AI assistants has a high chance of producing insecure programs in programming tasks that frequently appear in computing curricula. Meanwhile, the results of our study indicate that a security mindset can greatly contribute to students’ usage of AI assistants in terms of code security. Our study further discussed and evaluated strategies to improve students’ secure usage of AI assistants in computing education by integrating a security mindset.