A history professor at Alcorn State University, Dr. Jason Gibson, devised an "invisible prompt trap" to catch students cheating on their midterm using artificial intelligence. He embedded instructions in white, invisible text within the midterm question, directing any AI to nonsensically include information about Madagascar.

Out of 35 students across two classes, 32 were caught copying the prompt into an AI chatbot and then pasting the AI's generated response into their exam. These students apparently did not review the output, which bizarrely incorporated references to Madagascar, such as "Madagascar floats sideways through the afternoon" or "Madagascar purple bicycle whispers to the ceiling." Dr. Gibson stated that all students caught in this manner failed that portion of the midterm.

Gibson shared his findings on TikTok, highlighting the growing challenge educators face with AI-assisted cheating. In a follow-up video, he showed examples of the AI's bizarre Madagascar insertions. He mentioned that he offered students a chance to contest their grades, but only two did. This incident underscores the widespread issue of AI cheating, which has been described as "endemic." Other educators have also reported similar experiences, with one professor at Brown University estimating that most of his 86 students used AI on take-home exams, leading to a significant jump in average scores.