A Generative AI Engineer has developed an LLM application to answer questions about internal company policies. The engineer must ensure that the application doesn’t hallucinate or leak confidential data. Which approach should NOT be used to mitigate hallucination or confidential data leakage?