History professor Jason Gibson caught 32 students cheating on a midterm exam after they used AI to generate responses containing a hidden, nonsensical prompt about the nation of Madagascar.

Key Points

• Professor Jason Gibson identified 32 students across two classes who submitted AI-generated work without proofreading their responses.
• A hidden white-font instruction in the exam prompt required students to include the word "Madagascar" in a nonsensical context to detect automated submissions.
• Students failed to notice the prompt, resulting in exam answers that included bizarre phrases like "wore a toaster to a basketball game."
• Academic institutions, including Brown University and Princeton University, are reporting similar surges in AI-related cheating incidents.
• Princeton University recently abandoned its long-standing Honor Code tradition in favor of supervised exams to combat the rise of chatbot-assisted academic dishonesty.

Why it Matters

This incident highlights the growing challenge educators face in maintaining academic integrity as generative AI tools become easily accessible to students. It underscores a critical need for universities to develop new assessment standards and best practices to navigate the integration of AI in higher education.