A hypothetical situation could involve an AI-run customer care chatbot manipulated through a prompt containing destructive code. This code could grant unauthorized access to the server on which the chatbo