A Generative AI Engineer is building an LLM application that allows users to generate personalized birthday poems using their names. Considering the risk of malicious user inputs, which technique is the most effective for securing the application?