Asan Medical Center Implements Private AI Knowledge Search System on Closed Network

Search and Respond to Clinical Protocols Within Seconds, Including Sepsis Response Guidelines
Built on an On-Premises Model Using Only Internal Hospital Servers, Without External Cloud

On May 18, Asan Medical Center in Seoul announced that it has established and begun operating a private AI knowledge search system that runs in a closed network environment, completely separated from the external internet.

The system is built on an on-premises model, utilizing only the hospital’s internal servers and data. By not transmitting any data to external cloud services, the risk of patient medical records, test results, and other sensitive healthcare information being leaked outside is significantly reduced.

Medical staff can enter questions in natural language to quickly search for clinical guidelines and work regulations. The system is designed to allow users to access information based on internal hospital manuals—such as initial response protocols for suspected sepsis patients, antibiotic administration guidelines, emergency procedures for removing endotracheal tubes, and legal infectious disease reporting procedures—within seconds.

From a technical perspective, a vector database and Retrieval-Augmented Generation (RAG) technology have been implemented. Clinical guidelines and work documents within the hospital are stored in a format that enables semantic-based searches, and the AI generates responses based on these documents. This structure is intended to reduce the possibility of so-called “hallucinations,” where AI provides information that is not factually accurate.

To incorporate the latest external medical information, Asan Medical Center also plans to introduce an additional sandbox-type external search engine. Only anonymized questions, which cannot identify individual patients, will be sent outside the hospital’s network. The system was developed not by an external solution provider, but by the hospital’s own IT staff.

Kim Younghak, Head of the Digital Information Innovation Division at Asan Medical Center, stated, “This system addresses both security and AI utilization, demonstrating that AI can be used even in a closed network environment. We will continue to enhance the system to advance digital healthcare innovation while further reducing concerns about data leakage.”

This content was produced with the assistance of AI translation services.