China AI startup DeepSeek has launched an AI model that it claims is low-cost, high-performance, and comparable to OpenAI's, which has recently become a sensation. However, there have been security risks, suspected plagiarism of the US AI model, and various other controversies, raising concerns in the US, UK, France, Germany, Italy, Japan, and other countries, with many of them already taking countermeasures.
For example, the US Navy has banned the download of DeepSeek, and Italy's data protection authority has also ordered the restriction of DeepSeek's AI platform's access to Italian users' personal data. DeepSeek has also been removed from the Apple and Google app stores in Italy.
Taiwan's government agencies ban DeepSeek
Taiwan has also followed the international community's footsteps. The Ministry of Digital Affairs stated yesterday that, based on national cybersecurity considerations, it has specially warned government agencies and critical infrastructure to restrict the use of DeepSeek AI products to avoid the transmission of user-related data or information by products with cybersecurity concerns.
The Ministry of Digital Affairs pointed out that according to the Executive Yuan's reference guidelines for the use of generative AI, personnel in charge of business must not provide information involving confidential official duties, personal information, or information not publicly disclosed by the agency or institution to generative AI, nor ask generative AI questions that may involve confidential business or personal data. The DeepSeek AI service is a Chinese product, and its operation involves cross-border data transmission and information leakage, posing a threat to national cybersecurity.
The Ministry of Digital Affairs stated that to prevent internal information in government agencies from leaking and posing a risk under uncontrollable circumstances, Taiwan has clearly required central and local government agencies, public schools, state-owned enterprises, administrative legal persons, and venues for public activities or use to restrict the use of products that threaten national cybersecurity since the Executive Yuan implemented the "Principles for Restricting the Use of Products Threatening National Cybersecurity by Agencies" in 2019.
At the same time, the competent central authorities should supervise the critical infrastructure providers defined in the Cybersecurity Management Act and the government-funded foundations to reference the implementation.
The Ministry of Digital Affairs emphasized that during work hours, computers and network equipment must not be used for activities unrelated to official duties, and the cybersecurity protection of each agency is crucial. The Ministry of Digital Affairs will continue to monitor relevant technical developments and adjust cybersecurity policies in a timely manner to protect national information security.
Legislators urge: Not only ban, but also invest in AI development
Legislator Lin Chun-hsien stated that the Deepseek incident has become a major issue of international concern, and many countries have quickly taken countermeasures, so the Taiwanese government should quickly ban its use.
Lin Chun-hsien believes that not only Deepseek, but also other high-risk Chinese-made products and software, including TikTok, should be banned from use by the government. Especially since Deepseek has been confirmed as a high-cybersecurity-risk product, and the international community is taking countermeasures, Taiwan, which is on the frontline of the threat, should accelerate its response.
Legislator Wang Ting-yu stated that the US government has banned not only the public sector but also private use, and the US Navy has banned the use of this software. Taiwan's cybersecurity threat is no less than that of the US. As the US is a democratic country, and it has banned the use of Deepseek based on cybersecurity considerations, Taiwan, which is on the frontline of the threat, should be more cautious in its response.
Wang Ting-yu added that as for whether Deepseek has serious underestimated development costs, directly plagiarized the US's ChatGPT, has built-in self-censorship mechanisms, or has an information accuracy rate of only 17%, these are issues for the technology community to research. The government will continue to act as a gatekeeper for cybersecurity, and the public should also raise their awareness of self-information protection.
Legislator Chen Kuan-ting stated that he supports the government's ban on the use of DeepSeek, but Taiwan also needs to invest more resources in AI development. It cannot just ban without developing a system that is good enough for the public to use. He hopes that the private sector will quickly develop a better system that can connect with the international development of generative AI without affecting national security.
