hankyoreh

Lawmakers attend an urgent roundtable on Korea’s response to the DeepSeek shock and strategies for developing AI held at the National Assembly Members’ Office Building on Feb. 4, 2025. (Yonhap)

South Korean government agencies and private companies have begun restricting the use of the Chinese-developed generative artificial intelligence service DeepSeek. The moves have been prompted by concerns that sensitive work-related information could be leaked during use.

According to government sources, the Ministry of Trade, Industry and Energy (MOTIE), which is in charge of external trade policy, introduced measures Wednesday to restrict the use of DeepSeek on its computers with external internet access.

A day earlier, the Ministry of the Interior and Safety sent an official document to the MOTIE and other central government offices as well as metropolitan local governments with security guidelines urging caution when using generative AI services such as DeepSeek and ChatGPT.

In response, the MOTIE passed the guidelines along to its affiliated public institutions, while independently singling out DeepSeek as a generative AI service subject to access restrictions.

An official with the ministry explained that the move was made “in consideration of personal information protections and security.”

Among other central government offices, the Ministry of National Defense reportedly blocked DeepSeek access to internet-enabled computers involved in military duties, with the Ministry of Foreign Affairs also pursuing similar measures.

Sources said the Interior and Safety Ministry document transmitted the day before included instructions to refrain from entering personal information in generative AI and to avoid unconditionally trusting the content that AI services provide, but did not explicitly single out DeepSeek as a service that should not be used.

But the fact that institutions have independently begun restricting DeepSeek use comes in response to concerns that the service may be collecting too much user information amid its AI learning processes.

On Jan. 31, South Korea’s Personal Information Protection Commission sent a questionnaire to DeepSeek’s head office in China asking for confirmation on the types of personal information collected and the associated procedures and methods of processing and storage. No reply has yet been received.

Private companies have also begun moving to restrict DeepSeek use.

Kakao announced Wednesday that it had circulated a notice to its employees a day earlier calling on them to “refrain from using DeepSeek services for the purposes of company duties.”

The same day, mobile telecom company LG U+ banned the use of DeepSeek for duties and recommended that its employees refrain from using it even on their personal computers.

While Naver has not issued any separate announcements, it stated that it already has regulations prohibiting the use of generative AI for duties when the services in question store data on external servers, and that DeepSeek would not be an exception to this rule.

Korea Hydro & Nuclear Power, which is responsible for nuclear power technology, posted an intranet announcement on Saturday under the title “Use of the Chinese AI service DeepSeek prohibited.” The power corporation has also barred the use of ChatGPT for duties in the interest of nuclear power security.

Major corporations such as Samsung Electronics, SK, and LG Electronics have already developed and introduced their own internal generative AI programs and restricted the use of unauthorized external programs on company computers.

Revelations that DeepSeek was collecting various forms of data from its service’s users and storing them on its servers in China have sparked concerns about information leaks and security risks.

While other AI services such as ChatGPT also collect user account information and data provided to chatbots, DeepSeek was found to be collecting extensive information on everything up to and including keyboard input patterns and rhythms.

A telecommunications industry insider explained, “While there have been a lot of companies in the past that barred the use of generative AI for work-related purposes, this is the first time we’ve seen separate announcements [singling out DeepSeek].”

“I think a lot of it is that it’s still an unknown quantity in terms of security and safety,” they added.

A growing number of governments and public institutions around the world have also been prohibiting DeepSeek use. Australia, Japan, Taiwan and the US state of Texas have all banned its use on government-owned devices, while Italy has outright blocked downloading of the app.

By Park Jong-o, staff reporter; Jung Yu-gyung, staff reporter; Jang Soo-kyung, staff reporter; Shin Hyeong-cheol, staff reporter

Please direct questions or comments to [english@hani.co.kr]