I tried to manipulate the AI that “can’t hallucinate”: a prompt injection red-team test
Last month a startup called TypeSafe released a model called Jev, and the part that caught my attention was the claim that it can’t hallucinate. The first question…