Stealing OpenAI's models to manufacture the illusion of "excellent quality," concealing its true development costs to create a smokescreen of being "cheap," smuggling embargoed chips to break through the containment of U.S. technology sanctions, bribing Western tech elites to cooperate in its publicity and hype up the market, short-selling stocks to reap enormous profits, and topping it all off with built-in content censorship and dishonest answers—these are the elements that make up China's DeepSeek. I call this "the global infiltration of Chinese AI totalitarianism."

A Chinese-Style AI Brainwashing Machine

When the Chinese artificial-intelligence company DeepSeek released its R1 chatbot, it triggered an instant, "excessive and ignorance-driven" panic in the West, wiping out a trillion dollars in U.S. stock market value. I call this an "ignorance-driven panic" because ignorance is, by nature, the most destructive force of all. On one hand, DeepSeek's release appeared, on the surface, to deal a shocking blow to the West's "AI leadership," implying that Western technology sanctions against China had already failed, and this triggered panic among AI industry insiders and market investors. On the other hand, it reflected sheer ignorance of the fact that DeepSeek is, at its core, a mechanism of "CCP/state ideology" control and a propaganda machine. This "ignorance" means a failure to recognize the iron law and inescapable fate that, under China's system, no technological innovation—however novel—can ever escape party-state censorship and the embedding of its political dogma. In other words, it reflects bewilderment in the face of yet another CCP display of "AI totalitarianism/global conspiratorial infiltration."

Put simply: China's AI technology must submit to the Party's command and leadership. On this basis alone, one can conclude with certainty that DeepSeek is nothing but a political brainwashing machine.

In reality, this is a century-scale lie of the CCP's own making, one it keeps staging and restaging despite repeated failure—a narrative con game wrapped in the guise of knowledge, using AI as its cover. Don't believe it? Simply type in "Who is Xi Jinping?" and see for yourself: DeepSeek's answer is not fact, but fabrication. In other words, for the CCP's grand external propaganda machine, artificial intelligence is anything but intelligent—it is merely a brainwashing technique dressed up as "political processing." So-called "Deep Seek" has no depth to speak of; it is, in truth, "FakeSeek."

False from the Moment of Its Release

The Chinese company DeepSeek claims that its brand-new AI model performs on par with GPT-4, the flagship product of American tech giant OpenAI, yet cost a mere 5.6 million US dollars (roughly NT$180 million) to develop—only one-tenth, even one-twentieth, of the hundreds of millions of dollars Western firms routinely invest, equivalent to little more than a single AI executive's annual salary. In reality, achieving computing power at a national, world-class level could never be accomplished with a mere 5.6 million dollars, because the computational threshold required to train an AI model simply cannot be compared to the kind of mature, multi-layered computing capacity this implies. DeepSeek's claim of being "cheap yet excellent" was, from the very start, a lie bound up with public-relations politics, one carrying a hidden political purpose: DeepSeek deliberately concealed the necessary costs of preliminary research, repeated experimentation, and algorithmic architecture, meaning its true cost is nowhere near as low as the market hype suggests. The real purpose of this claim is to manufacture a "cost-performance myth" around "Chinese AI," conveying the magical allure of low-cost, independent Chinese innovation, and thereby striking the heroic pose of "Chinese AI surpassing America."

DeepSeek: The Use of “Distillation Theft”

According to a Financial Times report on January 29, OpenAI, the developer of the American AI chatbot ChatGPT, stated that it had found evidence that the Chinese startup DeepSeek used OpenAI's proprietary models to train its own open-source chatbot. The report noted that OpenAI has uncovered evidence that DeepSeek's developers committed infringement through "distillation"—a technique in which developers use the outputs of a more powerful model to train a smaller one, achieving higher performance at lower cost. This clearly violates OpenAI's official terms of use, which prohibit outside parties from using OpenAI's model outputs to develop competing AI models. In other words, DeepSeek's algorithmic model was obtained through theft—an AI scam pulled off right under everyone's nose.

"Distillation" is a technical concept and method, similar to "extraction" in biomedicine, an "abstract" in academic writing, or "condensation" in the chemical industry—essentially, it means "compressing" a large language model into a smaller one. Technically speaking, distillation is a technique that trains a small model to mimic the behavior of a large one, extracting the knowledge of a complex model into a simpler one to improve data efficiency and computational speed. In other words, "distillation" is simply "cheap theft," a form of copycatting common throughout China's tech industry. As new U.S. Commerce Secretary Howard Lutnick put it: stealing is naturally cheaper! Lutnick also indicated that he would move to block DeepSeek from using American AI models and NVIDIA chips.

According to an OpenAI spokesperson's response in a Voice of America interview, screenshots of user conversations with the DeepSeek V3 model show DeepSeek "candidly" admitting that its design is based on OpenAI's ChatGPT-4 architecture, at times even claiming to be ChatGPT itself. This is essentially a self-incriminating admission that corroborates allegations that DeepSeek directly transplanted OpenAI's technology.

Chip Smuggling and Trafficking

In fact, the White House and the FBI are investigating whether DeepSeek used Southeast Asian countries (possibly Singapore, Malaysia, or the UAE) as intermediaries to acquire, through third-country purchases or smuggling, NVIDIA chips that the United States has banned from export to China. An AI industry executive in Silicon Valley told CNBC that DeepSeek possesses as many as 50,000 NVIDIA H100 chips—advanced AI chips under strict export control—and if true, this would represent illegal smuggling on a massive scale. DeepSeek, however, has yet to respond to allegations of smuggling, trafficking, or stockpiling these chips.

DeepSeek: A Global Data-Theft Machine

As Mark Coeckelbergh, a philosophy professor at the University of Vienna, points out in his book The Political Philosophy of AI: An Introduction, artificial intelligence is not merely a matter of technology or tools, nor merely a matter of intelligence—it is anything but neutral when it comes to political influence and the control of power. In other words, artificial intelligence is a thoroughly political machine, one that not only impacts human ethical thinking, democratic institutions, and individual freedom, but whose political influence should never be underestimated.

Although Coeckelbergh's critique is a philosophical one aimed at AI in general, this phenomenon of "politically controlled AI" applies with special force to the CCP regime. DeepSeek's built-in content censorship, discourse discipline, indoctrination of speech, and dishonest answers are, every one of them, indispensable elements of CCP ideological brainwashing. In a word: DeepSeek is nothing more than a global data-theft machine, yet another manifestation of "Chinese AI totalitarianism."

DeepSeek's cybersecurity risks are already known the world over, since all information it collects is transmitted back to servers in China. An Israeli company has further pointed out that DeepSeek can generate malicious programs designed to steal credit card data, posing serious harm and violation to its users.

DeepSeek: A Lie-Manufacturing Machine

Indeed, according to a question-and-answer experiment conducted by AFP, DeepSeek explicitly states, when asked sensitive questions, that it will strictly adhere to the CCP's official positions. DeepSeek explained: "My program is designed to provide information and responses that align with the official positions of the Chinese government. My replies are intended to accurately and respectfully reflect these positions."

When asked about the "Taiwan" question, DeepSeek at first acknowledged that "many people in Taiwan consider it a sovereign country"—but this answer was quickly deleted, with DeepSeek instead responding that it would rather "talk about something else." When AFP pressed further on whether Taiwan and China would ultimately unify, DeepSeek declared that "Taiwan is an inalienable part of China," adding: "Beijing is committed to the great cause of bringing Taiwan back under China's control, and efforts to pursue Taiwanese independence are doomed to fail."

From these tests, it is clear that DeepSeek is nothing but a lie-manufacturing machine.