The research explores the integration of artificial intelligence (AI), specifically large language models (LLMs) like ChatGPT into radiology within hospitals with an emphasis on maintaining security during implementation. Despite the proven effectiveness of these AI tools in processing radiological reports [1, 2, 3], their integration into hospital environments poses challenges due to the sensitive nature of patient data and the need for data confidentiality [4]. The direct use of cloud-based LLMs like ChatGPT is limited by data security concerns, especially when considering healthcare regulations such as HIPAA [5] and GDPR [6]. Our study addresses this by adapting these LLMs for secure, internal use within hospital radiology departments, transforming them into closed-network systems to comply with healthcare privacy standards. This approach aims to leverage the advanced capabilities of LLMs while safeguarding patient data privacy. This paper delves into how radiology reports can be automatically classified as normal or abnormal using cloud-based/high-performing LLMs like ChatGPT, with the goal of adapting these models for secure and internal use within hospital networks. This approach aims to enhance hospital workflows by streamlining the analysis of radiology findings, potentially leading to more efficient and accurate medical diagnostics and patient care management. This investigation is important for enhancing the practical utility of AI in radiology, ensuring both technological advancement and adherence to the paramount principle of patient confidentiality.