Large language models (LLMs) are transforming healthcare through applications such as diagnosis, risk prediction, and treatment planning. While offering improved accuracy and efficiency, their use raises significant privacy and security concerns, including prompt injection, adversarial manipulation, and leakage of sensitive patient data. This chapter reviews key vulnerabilities, attack vectors, and their implications for medical practice. It also includes all medical fields branches and clinical use. We also summarize AI adoption across major medical fields, outlining benefits and challenges. Challenges and mitigation with LLMs in the medical field are illustrated and explained in detail. Some defense approaches with evaluation metrics such as privacy leakage rate, attack success rate, and computational overhead are concluded. Our findings highlight the need for robust, adaptive safeguards to ensure safe and ethical LLM deployment in healthcare. More researches in this direction are still open for academics and industries.