University Uses Hidden Prompt Injection (Font-Size 0 Text) in Assignments to Catch Students Pasting into LLMs — 50 Points
Hacker News·low signal
A first-year computer science course embeds invisible instructions in font-size 0 text within assignment prompts, so students who blindly paste the full prompt into an LLM trigger a detectable watermark in their submission. The 50-point, 25-comment HN thread represents a creative offensive approach to academic integrity — detecting AI-assisted cheating at the input stage rather than trying to identify AI-generated output after the fact. It's also a real-world example of prompt injection being used defensively rather than as an attack vector.