H/T China Fact Chasers YouTube channel
Recent discussions surrounding China’s AI chatbot, known as DeepSeek, have raised critical questions about its hardware, performance, data practices, and potential political motives. While the chatbot is promoted as a breakthrough technology available at little to no cost, a close look at its development and operations reveals a series of controversial issues.
DeepSeek is reported to run on Nvidia’s H100 chips—the high-performance components that are typically subject to strict export controls imposed by the United States. Investigations have noted that a significant quantity of these chips, allegedly numbering around 50,000, found their way to China via routes involving Singapore. Such transactions suggest that DeepSeek’s computational strength may stem from hardware obtained under questionable circumstances, challenging the narrative that the technology is built solely on domestic innovation.
In parallel with hardware concerns, performance metrics cast doubt on the chatbot’s effectiveness. An independent audit shared on social media claimed that DeepSeek delivered news and factual information with a relatively low accuracy rate when compared to Western AI counterparts like ChatGPT and Google Gemini. The audit reported frequent errors, including false claims and vague responses, suggesting that DeepSeek struggles to provide reliable outputs even as it is widely promoted.
Data privacy and censorship represent another layer of controversy. According to information provided on its own website, DeepSeek stores user data—including keystroke logs and other device information—on servers in China. This approach, combined with the chatbot’s built-in adherence to Chinese laws and socialist core values, means that its responses are tightly controlled. Such measures not only limit the range of acceptable topics and factual details but also align the chatbot’s output with state-sanctioned narratives. Users are effectively warned that sensitive topics may trigger self-censorship, making the technology a potential vehicle for propaganda.
Further complicating the picture is the method known as model distillation. Evidence from the transcript indicates that DeepSeek may have effectively “copied” aspects of established AI models by repeatedly querying them and mimicking their reasoning processes. This practice raises concerns about intellectual property and innovation; if DeepSeek is primarily learning from and imitating more advanced models, it may be inherently limited by its reliance on another’s groundwork.
Adding to the technological and operational issues, there are strong suggestions that the deep promotion of DeepSeek on social media was not organic but rather part of a coordinated influence campaign. Reports indicate that state-linked social media accounts and even foreign propagandist channels actively amplified positive messaging about DeepSeek. These efforts appeared designed to counterbalance critical perspectives, obscure the chatbot’s shortcomings, and potentially influence broader technology and financial markets by undermining confidence in Western AI solutions.
Collectively, the evidence paints a complex picture. DeepSeek is marketed as an innovative leap in artificial intelligence, yet it is burdened by controversies over hardware sourcing, underwhelming performance, invasive data collection, and a high degree of censorship. The integration of advanced yet possibly misappropriated hardware, coupled with aggressive state-backed messaging campaigns, suggests that the chatbot’s prominence may be as much about geopolitical strategy as it is about technological innovation.
Share this post