While global tech stocks last month were rattled by the launch of Chinese start-up DeepSeek’s artificial intelligence (AI) chatbot model DeepSeek R1, which claims to have been made at a low cost using less technologically advanced chips, the app has been banned by a growing number of countries over national security concerns, highlighting a new and dominant challenge Western democracies face with Chinese advanced technologies.
R1, which seemingly could match the performance of leading US AI models at a fraction of the cost and with free access, was launched on Jan. 20, the day US President Donald Trump was inaugurated, and a day before Trump announced a US$500 billion investment in Stargate, a US AI infrastructure project. The new app has been praised by Chinese officials and media as a symbol of the country’s breakthrough in AI tech, and was widely seen as an antagonistic response to the US’ tightened sanctions on exports of high-end chips and devices to China.
However, significant concerns regarding quality and cybersecurity have been raised about R1.
Industry analysts found that DeepSeek’s capital investment in hardware was at least US$1.6 billion, with operating costs estimated at US$944 million, much more than its claimed total cost of less than US$6 million. It even excluded money spent on prior research and ablation experiments on system architectures, algorithms and data.
Industry analyst firm SemiAnalysis found that DeepSeek has 50,000 high-end Nvidia Hopper graphics processing units, far more than the company initially reported. DeepSeek said it used old, lower-end Nvidia chips for its AI models. It was also proven that some of R1’s performance were based on existing models developed by Western firms and that it had stockpiled components before the US’ export curbs were implemented.
Furthermore, international cybersecurity firms found that R1 contains code capable of transmitting user data and sensitive information to China, and that its platform is vulnerable to data leakage and falsification. In less than a month since its launch, the app has been banned for use in government agencies in Italy, Australia, South Korea, Taiwan and Texas.
DeepSeek has also been criticized for providing biased data, especially those related to China. For example, it did not answer questions about Chinese President Xi Jinping (習近平) and did not give meaningful responses when asked about China’s human rights issues or issues such as the Tiananmen Square Massacre.
Similar to other Chinese apps, such as TikTok, and RedNote (Xiaohongshu, 小紅書), DeepSeek was developed to compete with the West and to expand China’s global market using opportunistic measures and penetrative pricing, which are accompanied by risks of cybersecurity leaks and data manipulation. Amid the tech competition between the US and China, the rise of R1 is a wake-up call and poses a new challenge to the US and Western tech giants, which have long led the industry, but have failed to thoroughly implement high-tech governance and export restrictions.
US experts have called for more collective actions to seal loopholes in the US and improve Western allies’ high-tech curbs. More comprehensive policies must also be developed to bolster AI outputs from democracies to lessen adoptions of Chinese models.
Taiwan in particular needs to speed up developing sovereign AI to reduce China’s impact, and utilize its strength in semiconductors and cooperate with more like-minded countries and institutions in strategic AI developments, such as in advanced generative AI applications and data centers. Taiwan could play a more crucial role in the democratic allies’ reindustrialization and in upholding their leading position over China in the AI race.
Speaking at the Copenhagen Democracy Summit on May 13, former president Tsai Ing-wen (蔡英文) said that democracies must remain united and that “Taiwan’s security is essential to regional stability and to defending democratic values amid mounting authoritarianism.” Earlier that day, Tsai had met with a group of Danish parliamentarians led by Danish Parliament Speaker Pia Kjaersgaard, who has visited Taiwan many times, most recently in November last year, when she met with President William Lai (賴清德) at the Presidential Office. Kjaersgaard had told Lai: “I can assure you that ... you can count on us. You can count on our support
Denmark has consistently defended Greenland in light of US President Donald Trump’s interests and has provided unwavering support to Ukraine during its war with Russia. Denmark can be proud of its clear support for peoples’ democratic right to determine their own future. However, this democratic ideal completely falls apart when it comes to Taiwan — and it raises important questions about Denmark’s commitment to supporting democracies. Taiwan lives under daily military threats from China, which seeks to take over Taiwan, by force if necessary — an annexation that only a very small minority in Taiwan supports. Denmark has given China a
Many local news media over the past week have reported on Internet personality Holger Chen’s (陳之漢) first visit to China between Tuesday last week and yesterday, as remarks he made during a live stream have sparked wide discussions and strong criticism across the Taiwan Strait. Chen, better known as Kuan Chang (館長), is a former gang member turned fitness celebrity and businessman. He is known for his live streams, which are full of foul-mouthed and hypermasculine commentary. He had previously spoken out against the Chinese Communist Party (CCP) and criticized Taiwanese who “enjoy the freedom in Taiwan, but want China’s money”
A high-school student surnamed Yang (楊) gained admissions to several prestigious medical schools recently. However, when Yang shared his “learning portfolio” on social media, he was caught exaggerating and even falsifying content, and his admissions were revoked. Now he has to take the “advanced subjects test” scheduled for next month. With his outstanding performance in the general scholastic ability test (GSAT), Yang successfully gained admissions to five prestigious medical schools. However, his university dreams have now been frustrated by the “flaws” in his learning portfolio. This is a wake-up call not only for students, but also teachers. Yang did make a big
RSS