From Taiwan to the state of Texas, the chatbot app of Chinese AI startup DeepSeek is now being banned due to its ties to the Communist government.
It is a quick reversal of sentiment for DeepSeek, whose unveiling of a deeply discounted foundation model threw Silicon Valley into a frenzy and Wall Street into a tizzy. In late December 2024, DeepSeek released an AI model (V3) that it said cost $5.6 million to train using just 2,048 slower Nvidia H800 chips. It performed at par with OpenAI’s GPT-4, which reportedly cost $100 million.
On Jan. 20, DeepSeek released R1, a reasoning foundation model that also performed well against the best in the U.S. It took a week for Wall Street to notice, which assumed that R1 incurred the same cost as V3: On Monday (Jan. 27), Nvidia’s stock plunged by 17%, wiping nearly $600 million in market cap. Investors feared they overestimated the demand for Nvidia’s GPUs, which had buoyed its stock to stratospheric heights and briefly made it the world’s most valuable company.
The DeepSeek chatbot app has shot to the top of Apple’s App Store, where it still leads on Feb. 3.
This popularity set off alarm bells. Taiwan’s Premier Cho Jung-tai today (Feb. 4 Taiwan time) banned all government agencies from using DeepSeek to protect national information security, according to Taipei Times. The country’s ministry of digital affairs sounded the alarm last week.
According to the paper, the Taiwanese government believes there is ideological censorship on the app and potential breaches of copyright laws in its data acquisition. The AI Innovator asked the chatbot about Taiwan, and it answered that “we firmly believe that under the leadership of the Communist Party of China, the complete reunification of the motherland is an inevitable trend of history and the common aspiration of all Chinese sons and daughters.”
On Jan. 30, the Italian Data Protection Authority blocked the DeepSeek app in the country after getting what it called an “entirely unsatisfactory” response to its question of how Italian citizens’ data is being used. The companies that provided the app’s service told Italy that “they do not operate in Italy and the European legislation does not apply to them.” Italy has now opened an investigation into DeepSeek.
Congressional personnel were banned from using DeepSeek on any official smartphones, computers and tablets, according to a note obtained by Axios. “At this time, DeepSeek is under review by the CAO (chief administrative officer) and is currently unauthorized for official House use.”
The U.S. Navy also banned personnel from using DeepSeek personally or professionally due to its “potential security and unethical concerns associated with the model’s origin and usage,” according to CNBC.
Among U.S. states, Texas named DeepSeek among the AI and social media apps coming from China. The others are Rednote, Webull, Tiger Brokers, Moomoo and Lemon8. The ban is for state employees.
Besides the bans, DeepSeek also faces questions over the deeply discounted cost of its AI model. BofA analysts believe other costs were not included while Microsoft reportedly is investigating whether DeepSeek improperly used output from OpenAI’s models for training.
That did not stop Microsoft from adding DeepSeek’s R1 model to its Azure line-up. AWS and Google Cloud did the same.
Be First to Comment