[Exclusive] Patient Information Targeted by China... Even Tertiary Hospitals Block DeepSeek

Samsung Medical Center and Kangbuk Samsung Hospital
Complete Separation of Internal and External Networks
Seoul National University Hospital, Asan Medical Center, and St. Mary's Hospital
DeepSeek Blocking Measures Imminent
Concerns Over Leakage of Patient and Medical Staff Information
"Possibility of Data Flowing to Chinese Government"

Major tertiary hospitals in South Korea are increasingly blocking access to the Chinese generative AI service 'DeepSeek' to prevent the leakage of sensitive data such as patient medical information.

According to the medical community on the 7th, it has been confirmed that generative AI services like DeepSeek are not accessible on the internal networks of hospitals such as Samsung Medical Center and Kangbuk Samsung Hospital. These hospitals have isolated their internal networks from external networks, effectively blocking any information leakage through AI. Seoul National University Hospital, Asan Medical Center, and Seoul St. Mary's Hospital are also expected to implement DeepSeek blocking measures on the same day. The National Cancer Center also blocked access to DeepSeek within the hospital the day before.

Asan Medical Center shared a notice with its staff the previous day stating, "The scope of information collected by DeepSeek is unclear. There are concerning indications that it collects additional information not gathered by other AIs," and added, "According to the China Data Security Law enacted in 2021, there is a possibility that the collected data could flow to the Chinese government."

An official from a major tertiary hospital said, "Hospitals generally have strict security measures, such as separating internal computer networks from external ones," but added, "They focused on the possibility of information leakage and took preemptive measures."

The medical field is expected to be one of the industries with the highest utilization of generative AI, including new drug development and hospital system improvements. However, it is also a sector where concerns about the leakage of personal information of patients and medical staff, as well as errors caused by the use of unverified information, are significant.

According to a study published in January this year by a research team at Asan Medical Center, the probability that generative AI such as ChatGPT and DeepSeek, and their core technology, large language models (LLMs), could bypass security systems and access sensitive personal information reached up to 80.8%. This was a joint study by Professor Younghak Kim of the Cardiology Department at Asan Medical Center and Dr. Taejun Jeon of the Big Data Research Center at Asan Life Sciences Institute.

Professor Younghak Kim said, "While the expected advancements from utilizing large language models in the medical field are significant, without enhanced data security, it could lead to serious personal information leaks. Since this field handles sensitive personal data, the importance of security is especially emphasized, and there is a need for independently operated, medical-specialized large language models."

Hospitals are repositories of vast amounts of data. The industry evaluates that the introduction of generative AI could reduce repetitive and simple tasks for medical staff and enable more efficient hospital operations. AI can learn from and diagnose numerous data such as X-rays, computed tomography (CT), and magnetic resonance imaging (MRI) to improve accuracy. It can also provide personalized treatment plans based on patients' medical data. However, if security is threatened, it could lead to the leakage of patients' sensitive personal information, raising ethical and legal risks.

Alongside hospitals, government ministries, local governments, and institutions are also taking measures to block DeepSeek or issuing cautionary guidelines. The pharmaceutical and bio industries have joined in the cautionary measures against DeepSeek. Daewoong Pharmaceutical decided to block access to DeepSeek on employees' work PCs starting the day before.

Yuhan Corporation does not use generative AI services such as DeepSeek or ChatGPT within the company for security reasons. Samsung Biologics, a bio company, has been blocking AI-related sites including ChatGPT since last year, considering the risk of information security leaks, and immediately blocked the Chinese DeepSeek after its release. Celltrion is preparing internal guidelines regarding the use of AI, including DeepSeek.