A hypothetical situation could involve an AI-run customer support chatbot manipulated via a prompt that contains destructive code. This code could grant unauthorized use of the server on which the chatbot operates, leading to sizeable protection breaches. Roughly twenty five several years ago