The vulnerability was discovered by cybersecurity and AI researcher David Kuszmar, who found that ChatGPT suffered from βtemporal confusion,β making it possible to put the LLM into a state where it did not know whether it was in the past, present, or future.