韓国省庁・警察、ディープシークへの接続遮断

【ソウルＡＦＰ＝時事】韓国政府の各省庁と警察は６日、中国の新興企業「ＤｅｅｐＳｅｅｋ（ディープシーク〈深度求索〉）」が開発した高性能の生成人工知能（ＡＩ）「Ｒ１」について、業務用コンピューターとの接続を遮断したと発表した。利用者情報の管理方法について、データ監視機関からの問い合わせに同社が応じなかったためとしている。（写真は、中国の新興企業ディープシーク〈深度求索〉のロゴ）
　韓国はフランスやイタリアなどと共に、データ取り扱いについての情報の開示をディープシークに書面で求めていたが、ディープシーク側はこの要請に返答しなかった。
　こうした状況を受けて複数の省庁は同日、生成ＡＩサービスを通じて機密情報が漏えいするのを防ぐため、接続を制限する措置を講じていることを確認した。
　国防省当局者はＡＦＰに対し、「ディープシークに対する遮断措置は、特にインターネットに接続された軍関連のパソコンを対象に実施している」と語った。
　韓国警察もＡＦＰに、ディープシークへのアクセスを遮断したと明らかにした。産業通商資源省は、すべてのパソコンについてアクセスが一時的に制限されたとしている。【翻訳編集ＡＦＰＢＢＮｅｗｓ】
〔ＡＦＰ＝時事〕（2025/02/06-17:04）

South Korean ministries and police said Thursday they were blocking DeepSeek's access to their computers, after the Chinese AI startup did not respond to a data watchdog request about how it manages user information.
DeepSeek launched its R1 chatbot last month, claiming it matches the capacity of artificial intelligence pacesetters in the United States for a fraction of the investment, upending the global industry.
South Korea, along with countries such as France and Italy, have asked questions about DeepSeek's data practices, submitting a written request for information about how the company handles user information.
But after DeepSeek failed to respond to an enquiry from South Korea's data watchdog, a slew of ministries confirmed Thursday they were taking steps to limit access to prevent potential leaks of sensitive information through generative AI services.
Blocking measures for DeepSeek have been implemented specifically for military work-related PCs with Internet, a defence ministry official told AFP.
The ministry, which oversees active-duty soldiers deployed against the nuclear-armed North, has also reiterated the security precautions regarding the use of generative AI for each unit and soldier, taking into account security and technical concerns, it added.
South Korea's police told AFP they had also blocked access to DeepSeek, while the trade ministry said that access had been temporarily restricted on all its PCs.
The trade, finance, unification and foreign ministries also all said they had blocked the app or had taken unspecified measures.
- Bans 'not excessive' -
Last week, Italy launched an investigation into DeepSeek's R1 model and blocked it from processing Italian users' data.
Australia has also banned DeepSeek from all government devices on the advice of security agencies.
Kim Jong-hwa, a professor at Cheju Halla University's artificial intelligence department, told AFP that amid growing rivalry between the United States and China he suspected political factors could be influencing the reaction to DeepSeek -- but said bans were still justified.
From a technical standpoint, AI models like ChatGPT also face numerous security-related issues that have not yet been fully addressed, he said.
Given that China operates under a communist regime, I question whether they consider security issues as much as OpenAI does when developing innovative technologies, he said.
We cannot currently assess how much attention has been paid to security concerns by DeepSeek when developing its chatbot. Therefore, I believe that taking proactive measures is not too excessive.
Beijing on Thursday hit back against the ban, insisting the Chinese government will never require enterprises or individuals to illegally collect or store data.
China has always opposed the generalisation of national security and the politicisation of economic, trade and technological issues, foreign ministry spokesman Guo Jiakun said.
Beijing would also firmly safeguard the legitimate rights and interests of Chinese enterprises, Guo vowed.
- 'Complex competition' -
DeepSeek says it uses less-advanced H800 chips -- permitted for sale to China until 2023 under US export controls -- to power its large learning model.
South Korean chip giants Samsung Electronics and SK hynix are key suppliers of advanced chips used in AI servers.
The government announced on Wednesday an additional 34 trillion won ($23.5 billion) investment in semiconductors and high-tech industries, with the country's acting president urging Korean tech companies to stay flexible.
Recently, a Chinese company unveiled the AI model DeepSeek R1, which offers high performance at a low cost, making a fresh impact in the market, acting President Choi Sang-mok said Wednesday.
The global AI competition may evolve from a simple infrastructure scale-up rivalry to a more complex competition that includes software capabilities and other factors.