The DeepSeek dilemma: navigating innovation and security concerns in Korea’s healthcare industry < Special < Article

In the rapidly evolving landscape of healthcare technology, Korea finds itself at a critical crossroads following the emergence of DeepSeek, China’s latest breakthrough in generative artificial intelligence.

This development has ignited a nationwide debate on balancing technological advancement with data security, particularly in the sensitive healthcare sector.

The controversy escalated when Korea’s National Intelligence Service (NIS) released a detailed technical analysis of DeepSeek last Sunday.

The agency’s findings revealed security vulnerabilities that extend far beyond typical data privacy concerns. According to the NIS, DeepSeek’s architecture includes mechanisms for recording identifiable keystroke patterns and maintaining communication channels with Chinese servers via domains like volceapplog.com, potentially enabling unauthorized transmission of user conversations and sensitive data.

What makes DeepSeek particularly problematic for healthcare institutions is its unrestricted approach to data handling.

Unlike its competitors in the generative AI space, NIS stressed that DeepSeek lacks fundamental controls to prevent the absorption of user input data into its training datasets.

“This means that sensitive medical information, patient data, and proprietary research could potentially be incorporated into the AI’s learning model without user consent or knowledge,” the NIS said. “Also, DeepSeek’s terms of service permit unlimited sharing of user information with advertisers, with no specified data retention period – a particularly alarming revelation for healthcare providers handling sensitive patient information.”

Korean healthcare industry response

The response from Korea’s healthcare sector has been swift and decisive so far.

Even before the NIS report, the Ministry of Health and Welfare (MOHW) and the Ministry of Food and Drug Safety decisively implemented a temporary block on DeepSeek access last Thursday, citing concerns over potential data leaks.

The MOHW also instructed its affiliated organizations to implement similar blocking measures, marking a significant escalation in the government’s response to the security concerns surrounding Chinese AI technology.

This protective measure was quickly adopted across the industry. Samsung Biologics immediately banned DeepSeek access and extended restrictions to all AI-related sites, including ChatGPT.

Daewoong Pharmaceutical also blocked DeepSeek access for all employee workstations as of Feb. 6, and Yuhan Corporation expanded their restrictions to encompass both DeepSeek and ChatGPT.

LG Chemical had already maintained similar restrictions on external AI site access, demonstrating the industry’s growing awareness of security concerns even before the NIS warning.

SK biopharmaceuticals joined this industry-wide response by implementing comprehensive measures. The company is not only reviewing access restrictions to DeepSeek but is also developing broader guidelines covering the use of external generative AI services and strengthening data security protocols.

Kakao Healthcare has recently issued guidance to its employees, advising against the use of DeepSeek for internal business purposes, aligning with their AI utilization policies.

Some companies that have invested in their own AI capabilities, such as medical AI imaging firm Lunit, have remained independent of external AI models like DeepSeek, relying instead on proprietary technology.

However, the situation has become more complex for companies that are maintaining significant operations in China.

Organizations like Hanmi Pharmaceutical face the challenging task of balancing security concerns with business necessities.

These companies have adopted a nuanced approach, allowing their Chinese subsidiaries to maintain independent security protocols while implementing stricter measures in their Korean operations.

Hospitals with vast amounts of patient data have also begun checking for DeepSeek restrictions.

Samsung Medical Center and Kangbuk Samsung Hospital have blocked the use of generative AI, including DeepSeek, by separating their internal and external networks. Seoul National University Hospital, Asan Medical Center, and Seoul St. Mary’s Hospital are planning to implement DeepSeek restrictions. The National Cancer Center has also blocked access to DeepSeek within the hospital.

New business models arise

Despite these security challenges, the potential impact of generative AI in healthcare remains substantial.

Market projections from Global Market Insights suggest that the healthcare generative AI sector will experience remarkable growth, expanding from $1.8 billion in 2023 to $22.1 billion by 2032, representing an impressive annual growth rate of 32.6 percent.

The McKinsey Global Institute’s analysis provides even more optimistic figures, suggesting potential markets worth up to $28 billion in research and early-stage drug discovery, and an additional $25 billion in clinical development.

KoreaBIO has also stated that generative AI will play a crucial role in drug development.

“Generative AI models support drug development by generating new small molecules, nucleic acid sequences, and proteins with desired structures and functions,” the biotech industry group recently reported. “By analyzing successful drug chemical structures and testing variations, it can produce potential drug candidates at a faster rate than traditional drug development methods.”

KoreaBIO further elaborated that AI can help identify new targets for drug development accurately, leading to the development of more effective treatments as it can be utilized in designing and modifying complex clinical trials.

Korean healthcare companies are actively pursuing these opportunities through various innovative approaches.

Samsung Biologics invested in Generate Biomedicines, a U.S. biotech venture company possessing generative AI-powered protein design technology, at the end of last year.

SK biopharmaceuticals has recently established a joint venture with Eurofarma, one of South America’s largest pharmaceutical companies, to develop an AI-based epilepsy management platform.

The company has also been proactive in partnering with AI drug candidate discovery firms to find the next blockbuster drug.

JW Pharmaceutical continues to advance its drug candidate discovery efforts through its proprietary AI platform, J-Wave, while emphasizing that its internal server-based approach provides enhanced security compared to external AI services.

Additional concerns beyond security

However, the integration of generative AI in healthcare faces challenges beyond security concerns. KoreaBIO has highlighted the critical issue of AI hallucination – instances where AI systems generate plausible but incorrect information.

This phenomenon is particularly concerning in healthcare settings, where accuracy is paramount.

The industry group emphasizes that AI-generated results must undergo thorough expert review before being shared with patients, as inaccurate AI diagnoses and treatments could threaten personal health and create confusion regarding medical decision-making responsibility.

“Specific regulatory measures must be established for generative AI used in clinical settings,” KoreaBIO said.