The article discusses the application of generative AI and large language models (LLMs) in healthcare, focusing on their potential and challenges. Generative AI, designed to create new content from trained parameters, has gained significant attention, particularly with the launch of ChatGPT by OpenAI. In healthcare, these models are being applied to electronic medical records (EMRs) to improve predictive performance, simplify model development, and reduce costs. However, concerns about data privacy, generalizability, and the potential for "hallucination" responses remain. Wornow et al. conducted a review of 84 foundation models trained on clinical structured text data from EMRs, distinguishing between clinical language models and EMR models. They found that while these models can enhance predictive accuracy, they are limited by small datasets and data privacy issues. To address these limitations, Wornow et al. propose an evaluation framework with six criteria: predictive performance, data labeling, model deployment, emergent clinical applications, multimodality, and novel human-AI interfaces. The article also highlights recent advancements, such as Microsoft's partnership with Epic EHR and Oracle Cerner's integration of generative AI into their EHR systems. These developments underscore the need for a robust evaluation framework to assess the clinical value of these models. Additionally, the article emphasizes the importance of leadership, incentives, and regulation to ensure the responsible and effective implementation of generative AI in healthcare. Leadership is crucial for model development, validation, and implementation, while continued regulation is necessary to balance the interests of various stakeholders. Payer incentives are also essential for widespread adoption. The article concludes by calling for a comprehensive approach to address the challenges and opportunities presented by generative AI in healthcare.The article discusses the application of generative AI and large language models (LLMs) in healthcare, focusing on their potential and challenges. Generative AI, designed to create new content from trained parameters, has gained significant attention, particularly with the launch of ChatGPT by OpenAI. In healthcare, these models are being applied to electronic medical records (EMRs) to improve predictive performance, simplify model development, and reduce costs. However, concerns about data privacy, generalizability, and the potential for "hallucination" responses remain. Wornow et al. conducted a review of 84 foundation models trained on clinical structured text data from EMRs, distinguishing between clinical language models and EMR models. They found that while these models can enhance predictive accuracy, they are limited by small datasets and data privacy issues. To address these limitations, Wornow et al. propose an evaluation framework with six criteria: predictive performance, data labeling, model deployment, emergent clinical applications, multimodality, and novel human-AI interfaces. The article also highlights recent advancements, such as Microsoft's partnership with Epic EHR and Oracle Cerner's integration of generative AI into their EHR systems. These developments underscore the need for a robust evaluation framework to assess the clinical value of these models. Additionally, the article emphasizes the importance of leadership, incentives, and regulation to ensure the responsible and effective implementation of generative AI in healthcare. Leadership is crucial for model development, validation, and implementation, while continued regulation is necessary to balance the interests of various stakeholders. Payer incentives are also essential for widespread adoption. The article concludes by calling for a comprehensive approach to address the challenges and opportunities presented by generative AI in healthcare.