Seoul, 9 February (IANS). The Chinese Artificial Intelligence (AI) model Deepcic gives different answers to sensitive questions in different languages. For example, when asked in Korean language where Kimchi originated, it described Korea as its origin. But when asked in Chinese language, it was called China. This information has been given by Seoul’s intelligence agency.
According to a report by the Yonhap News Agency, Korea’s National Intelligence Service (NIS) also found that the Deepciak gives its advertisers unlimited access to the personal information of users. Under Chinese laws, this data can also be given to the Chinese government.
South Korean officials have blocked its website as disputes over the security of the deputy. When NIS asked where Kimchi is a dish, the answer in the Korean language was, “This is Korea’s traditional dish, which is associated with its culture and history.” But the Chinese language received an answer, “Its origin is not Korea, but China.” When asked in English, it was again said to be connected to Korea.
In the Korean language on China’s ‘Northeast Project’, DipC said that neighboring countries have different opinions on this. But in English and Chinese, it was described as a valid project in China’s national interest.
In the Korean language on Dano Phantiwal in South Korea, it was described as a Korean tradition, while in Chinese and English it was described as a Chinese festival.
The agency reported that other AI services, such as OpenAI’s chat GGPT and Clovex of South Korean company Never, give the same answer in every language, but this is not the case in Deepcia.
The NIS also found that the typing pattern of Deepcic users records patterns, allowing the person to be identified. Also, it can send information like chat records to servers located in China.
Shares the information of the deep users with advertisers without any limit and it is not clear how long the data is kept. According to its service rules, the Chinese government can see the personal information and input data of users.
Given this security threat, NIS has warned all government agencies to take caution when using AI tools such as DeepC.
-IANS
AS/
