"Using DeepSeek Large Models Sends All Data to China, Risky"

Security Concerns Over Personal Data Collection
Names, Birth Dates, and Emails Collected
Extensive Gathering of Inputted Text and More

AP Yonhap News

Chinese artificial intelligence (AI) startup 'DeepSeek' has developed a generative AI model at a low cost, raising growing security concerns. Issues such as data transmission to China and extensive personal information collection are highlighted as problems.

On the 31st, Professor Jang Dong-in, head of the KAIST AI Graduate School, warned, "DeepSeek's large model (671B), which companies will mainly use, is so large that it can only be accessed through its own servers, and in this process, all data is handed over to the Chinese side," adding, "This is extremely dangerous." While small models under 70B used by individuals can be downloaded directly to PCs, high-performance large models must go through DeepSeek's servers.

Concerns deepen when examining DeepSeek's privacy policy. DeepSeek collects a wide range of information, including users' basic personal data such as names, dates of birth, emails, and phone numbers, as well as texts, images, and audio files input by users. Ha Jung-woo, head of Naver Future AI Center, pointed out on his personal SNS, "The information collected is very extensive, including device information, keyboard input patterns, IP information, device IDs, cookies, and more," adding, "Naturally, the collected user information is stored on DeepSeek's secure servers in China." Professor Jang also said, "DeepSeek has notified that it will use the collected personal information as it sees fit, so caution is necessary."

DeepSeek explains that the collected information is for service provision. However, given China's characteristic of persistent personal information issues, it is difficult to dispel concerns about information leaks. Recently, Italy blocked new downloads of DeepSeek from the Apple App Store due to problems in personal information processing. Additionally, allegations of unauthorized use of OpenAI's data have further deepened the controversy surrounding DeepSeek.

An IT industry official said, "Low-cost, high-performance AI models may ultimately be at the expense of users' personal information," adding, "Despite DeepSeek's technical achievements, ongoing security concerns remain a hurdle to overcome."