This study explores the significant challenges posed by advanced large language models (LLMs) such as ChatGPT, particularly within educational contexts focused on computer science and programming education. These state-of-the-art LLMs have the capability to generate solutions for routine exercises designed to enhance students’ analytical and programming abilities. However, the ease with which AI can produce programming solutions threatens educational processes and skill development by potentially encouraging reliance on AI-generated answers rather than fostering independent problem-solving. Our study offers a theoretical illustration through a formal decision-making model, demonstrating how the ability to verify the originality of work and detect plagiarism can influence students’ motivation for independent work. This study suggests a collaborative framework involving computer science educators and AI developers to equip evaluators with tools capable of distinguishing between code authored by students and code generated by LLMs. This collaboration includes embedding unique AI engine signatures into the generated code, with specific details determined jointly by the academic team and the AI company. To facilitate this embedding, we introduce a novel application of steganography for watermarking AI-generated code. By implementing this multifaceted approach and harnessing such technologies through partnerships among educators, course administrators, and AI experts, we aim to maintain the integrity of programming education in an era increasingly dominated by advanced LLMs.