2026 ASEE Annual Conference & Exposition

Large Language Models (LLMs) have rapidly become tools of fascination for students and professionals alike, offering a sense of effortless progress, almost like a time machine that transports users from an initial idea to a polished state of completion within moments. Their convenience, however, often obscures underlying risks: in the rush to achieve quick results, users may unknowingly share sensitive, private, or protected information with these systems or generate content that compromises privacy and security. While LLMs offer immense potential for creativity, productivity, and problem-solving, their secure and responsible use requires awareness and caution.

We redesigned a CS-2 Problem solving course to integrate AI, this paper discusses how we addressed ethical, privacy, and security implications of using LLMs throughout the course. During Week 2, students explored the risks and benefits of LLMs and engaged with materials on LLM privacy and safety, how the government and individual organizations are addressing privacy and security needs. In Week 3, students completed the CS Personality Bot Interaction and Ethics Assignment, where they interacted with Character AI and reflected on ethical concerns through assigned readings from CBS News and NPR about lawsuits involving minors using chatbots. They connected these discussions to the ACM Code of Ethics, focusing on professional responsibility, respect for user privacy, and harm prevention. By Week 5, the focus shifted to Software Engineering, where students analyzed how LLMs intersect with security practices. Content on security concerns, secure coding practices, how LLMs can assist with security, and security risks of LLMs, encourage students to consider both the potential and the vulnerabilities of integrating AI tools into software development. Collectively, this multi-week intervention fostered ethical and security awareness while promoting responsible AI use in professional computing contexts.

As part of the intervention, we conducted a pre and post course survey which included an important question: “What are the security concerns around LLMs?” This question was designed to capture students’ baseline awareness before the instructional modules and measure conceptual growth afterward. The qualitative analysis of the open ended responses showed an increase in the students' recognition of key LLM-related security issues, particularly in areas such as data leakage, personal information gathering, and malicious use. We also noted the emergence of new themes in their answers such as “hallucination”, “privacy issues” and “prompt injection attacks.” These additions indicate that students developed a more nuanced and technically informed understanding of LLM-related risks, extending beyond general privacy and misuse concerns to encompass deeper issues of model behavior and reliability. This shift suggests that the course intervention effectively deepened students’ understanding of real-world AI security and ethical risks. Our results demonstrate that this lightweight approach of highlighting potential LLM security risks in an introductory computing course can increase students' awareness of the need for responsible usage.