A hypothetical scenario could require an AI-powered customer care chatbot manipulated via a prompt that contains malicious code. This code could grant unauthorized usage of the server on which the chatbot