DeepSeek from China that also provides lottery numbers... banned overseas [AK Radio]

News of Winning Lottery Numbers in China Draws Attention
Development Cost Only 1/18th of ChatGPT
Weak Personal Information Protection... Commercialization Challenging

China's generative artificial intelligence (AI) 'DeepSeek' has recently attracted global attention by making headlines with a lottery winning case. News that a man from Hunan Province, China, won the lottery using numbers recommended by DeepSeek rapidly spread through Chinese social media, intensifying the DeepSeek craze.

The man reportedly won with one of the five number combinations recommended by DeepSeek. Although the actual prize was only about 5 yuan (approximately 1,000 KRW), the news spread nationwide in China, leading many Chinese people to inquire about lottery numbers from DeepSeek. As server overload caused by the surge in access became severe, Chinese authorities intervened, stating that the possibility of predicting lottery winning numbers through AI is almost nonexistent. Experts also warned against gambling, emphasizing that lottery draws are random and cannot be solved through big data or formulas.

In terms of technical achievements, DeepSeek is not particularly superior compared to existing generative AIs like ChatGPT, but it has shocked the AI industry with its remarkably low development cost. DeepSeek's disclosed development cost is about 5.6 million USD (approximately 8 billion KRW), which is only about 1/18th of OpenAI's ChatGPT initial development cost of 100 million USD (approximately 144 billion KRW). While most countries plan to invest billions of dollars in AI development, DeepSeek's low-cost development is causing a significant ripple effect across the entire AI market.

However, the AI industry is raising doubts about the reliability of the development cost disclosed by DeepSeek. Major research firms estimate that the actual development cost likely reached at least 500 million USD. It is analyzed that the Chinese government may have employed indirect subsidies or various measures to reduce corporate burdens. Especially, since advanced AI semiconductors will be required for future infrastructure construction and upgrades, there are forecasts that additional costs could reach astronomical levels.

Reuters·Yonhap News

Behind DeepSeek's rapid growth, concerns about personal information protection are also growing. In South Korea, the app download has been suspended since the 15th following a warning from the Personal Information Protection Commission. Major advanced countries, including the United States, have also completely banned the use of DeepSeek by public institutions. The main reason is concerns over personal information leakage.

To use DeepSeek's service, users must provide personal information such as date of birth, name, phone number, and email address, all of which are stored on Chinese servers. Notably, Article 7 of China's National Intelligence Law stipulates that all organizations and citizens must cooperate with government intelligence activities, raising concerns that the collected personal information could be used without authorization by the Chinese government.

Security vulnerabilities are also pointed out as a major issue. DeepSeek is evaluated to have a generally lower security level compared to other AI services. In particular, there are concerns that questions or information entered by users could be used as a means of technology leakage. For example, information input during document summarization or project document work could potentially be leaked to the Chinese government. The United States has been especially sensitive to this issue, given past attempts by China to steal military secrets and space technology.

On the 6th, access to the DeepSeek site was blocked on a PC at the Government Seoul Office in Jongno-gu, Seoul. Photo by Yonhap News.

South Korea is taking a cautious stance on the DeepSeek issue amid the US-China rivalry. The government is implementing conditional restrictions based on resolving security issues, which differs from the full bans adopted by the US, Europe, and Taiwan. Experts worry that this issue could become a new source of conflict in Korea-China relations, similar to the THAAD deployment incident.

As the US and China are expected to shift into negotiation mode to resolve trade disputes in 2024, there is a possibility that DeepSeek could emerge as a new conflict factor. In particular, South Korea is likely to face deepening deliberations over the scope of private sector use, separate from restrictions on public sector use. Despite the advantage of cost efficiency, considering the overall trend of Western countries suspending use, immediate full-scale adoption seems unlikely.

Establishing guidelines for private companies' use of DeepSeek is also emerging as an urgent task. It is necessary to find a balance between national challenges of personal information protection and technology security and corporate benefits of cost reduction. Experts advise that, in the long term, strategic approaches are needed to secure competitiveness in the global AI market alongside investments in domestic AI technology development.