Key points:
- U.S. intelligence agencies accused Chinese AI developers of illegal “distillation” of American models from U.S. frontier models.
- China’s Foreign Ministry and embassy officials rejected the accusations as politically motivated smears.
- DeepSeek achieved technical breakthrough with minimal hardware, challenging narrative of U.S. AI superiority.
- Evidence suggests DeepSeek’s design is a simple database file requiring Western software to run.
- Panic response included stock crash and apparent cyberattack on DeepSeek servers.
- Debate raises questions about free speech, technological competition, and the future of global innovation.
The distillation gambit: A convenient excuse for missed opportunity
On Tuesday, the Federal Bureau of Investigation, the National Security Agency, and the Cybersecurity and Infrastructure Security Agency issued a joint cybersecurity advisory accusing China-based artificial intelligence companies including DeepSeek, Alibaba, Moonshot AI, MiniMax, StepFun, and Z.AI of engaging in “aggressive, malicious, and targeted distillation activities at an industrial scale” that extract “restricted proprietary functionalities and capabilities of U.S. frontier AI models.” The agencies allege that since late 2024, these firms have extracted billions of tokens from models such as Anthropic’s Claude, OpenAI’s GPT, Google’s Gemini, and SpaceXAI’s Grok, routing requests through multiple pathways to bypass geographic restrictions and violate terms of service.
The timing is telling. This accusation comes not on the heels of a Chinese intelligence failure, but immediately after DeepSeek demonstrated that a lean, efficient model could compete with resource-guzzling American systems. The method of attack is described as “distillation,” a legitimate and widely used research technique where a smaller, less capable model is trained on outputs from a larger, more effective one. While the intelligence agencies claim this constitutes unauthorized extraction, the reality is that frontier models are designed to interact with users, and any interaction produces outputs that can be used for training. The U.S. companies themselves have openly marketed their APIs for exactly this purpose.
In Beijing, the Ministry of Foreign Affairs spokesperson Mao Ning told a regular press conference that “China’s AI development is the result of high-level technological self-reliance and strength,” adding that both nations should “strengthen cooperation to promote AI development that is open, inclusive, universally beneficial and oriented toward the common good.” Liu Chang, spokesperson for the Chinese Embassy in Washington, called the accusations a “deliberate attack on China’s development and progress in the AI industry,” stating that the development of artificial intelligence in China is “based on strengthening the country’s self-reliance and scientific and technological capabilities.”
The hollow crown: When the emperor has no clothes
The deeper story, however, is not about accusations but about the systemic weakness those accusations are designed to conceal. When a user downloads DeepSeek, the file is an unprotected GGUF format, a collection of numerical weights that represent language without any executable code. To even run the model, one needs additional software developed primarily in the United States or Western Europe, such as Ollama or LM Studio. This simple technical fact undermines the entire narrative of China stealing secrets or putting spyware into American hardware. Instead, it points to a more straightforward explanation: Chinese engineers, benefiting from a higher annual graduation rate for science and technology workers than the United States, built a competitive product using openly available techniques and cheaper hardware.
Treasury Secretary Scott Bessent told an event at Southern Methodist University that “the Chinese distill our models and they can never get ahead of us.” Yet history suggests otherwise. The United States has a track record of dismissing foreign innovation as a hoax, as seen with cold fusion, now called Low Energy Nuclear Reactions, which was ridiculed by the American scientific establishment. Despite hundreds of replications by laboratories worldwide, research has progressed more actively in Russia and Japan. Similarly, the recent discovery of room temperature superconductors known as LK 99 was declared a hoax by Nature under what critics call pressure from a “scientific mafia.” South Korea has continued replicating and publishing findings.
The pattern suggests a structural resistance within American institutions to acknowledging competitive breakthroughs from outside the established order. When a small team of Chinese developers achieves something that Silicon Valley said required billions of dollars and years of chip development, the response is not to celebrate human advancement but to attack the messenger. The cyberattack against DeepSeek following the stock market crash on Monday appeared conveniently timed to discredit the platform, even as U.S. officials claimed Chinese developers were the aggressors.
Sources include:
Please contact us for more information.
