South Korea Joins International Effort to Block Chinese AI Amid Privacy Concerns

South Korea plans to block access to Chinese AI DeepSeek over data privacy concerns. This move echoes actions by Taiwan, the US, Italy, and Australia, all worried about data security and transparency issues related to the AI service's operations.

In a significant policy shift, the South Korean finance ministry has announced its intention to block access to the Chinese artificial intelligence service DeepSeek, citing data privacy concerns. The decision, as reported by Yonhap News Agency on Thursday, follows multiple domestic and global technical concerns surrounding the service.

Recently, the country's defense, foreign, and trade ministries have restricted access to DeepSeek, and the unification ministry is likely to follow suit. According to an official, since 2023, undisclosed official data has been barred from all generative AI models at the behest of national intelligence and interior ministries, signaling increased scrutiny over data protection.

The Ministry of Environment has also joined the crackdown, blocking DeepSeek from internet-connected government computers as of Thursday morning. This precautionary measure was taken after intelligence authorities highlighted uncertainties about DeepSeek's personal information collection processes.

Internationally, the apprehension about DeepSeek is growing. Taiwan's Ministry of Digital Affairs (MODA) earlier banned its use, fearing potential data leaks to Beijing, while Italy's data protection authority has cited transparency concerns as its basis for blocking the AI model.

The implications of DeepSeek were also highlighted in the United States, where Howard Lutnick, a nominee for US Commerce Secretary, accused the AI service of exploiting US technology and circumventing export controls to advance its models. Lutnick warned of China's technological ambitions and stressed the need for robust measures to protect national interests. Similar concerns have been voiced by Australian officials.

(With inputs from agencies.)