Concerns are mounting in the U.S. over Chinese artificial intelligence (AI) companies rapidly catching up to the performance of cutting-edge American AI models through a technique called distillation. This development is raising alarms about potential threats to U.S. AI competitiveness and has ignited a heated debate over regulations surrounding open-source AI.
According to industry sources on Thursday, the Wall Street Journal (WSJ) recently reported, citing U.S. government and industry officials, that there is growing apprehension regarding Chinese AI firms swiftly enhancing their competitiveness through distillation techniques that leverage American AI models.
Distillation is a technology that improves the performance of smaller models by learning from the responses of larger, high-performing AI models. This method enables developers to create high-performance AI with relatively fewer computing resources and costs, making it widely adopted in the AI industry recently.
In the U.S., there is particular concern that Chinese companies are effectively absorbing advanced AI technologies from American firms through this approach at an alarming rate.
The WSJ reported that small model developers, including several Chinese companies, are using distillation technology to mimic the capabilities of America’s top models. Experts emphasizing AI safety worry that if open models evolve and do not adhere to the same regulations as U.S. companies like Anthropic or OpenAI, threats such as cyber or biological weapons could emerge.
Chinese AI startup Moonshot AI recently sent shockwaves through the industry by unveiling its top-tier AI model, Kimi K3. Designed as an open model, Kimi K3 demonstrated performance levels competitive with the latest U.S. models in some benchmarks. Moonshot AI plans to release all weights of Kimi K3 in the near future.
In response, the WSJ noted that executives from major U.S. AI companies expressed strong concern, using forceful language.
Dean Ball, OpenAI’s Head of Strategic Futures, criticized the potential dominance of open-source models, warning that such a scenario could lead to complete AI communism, where technology becomes a state-led digital public good rather than a market commodity. He described this as a dystopian hellscape.
Dario Amodei, Chief Executive Officer (CEO) of Anthropic, also consistently warns about the risks to national security and potential cyberterrorism that could arise if AI models with robust cybersecurity features are indiscriminately released as open-source.
The WSJ analyzes that the heightened sense of crisis among these leading U.S. companies is directly linked to financial realities.
For big tech firms that must invest billions of dollars in computing costs to develop next-generation technologies, the availability of Chinese models for free poses a significant threat to their revenue models, especially as they approach initial public offerings (IPOs) and invest trillions of dollars in AI infrastructure.
Reflecting this concern, security officials within the Trump administration and some lawmakers are considering various pressure tactics, including blacklisting Chinese AI companies, issuing security warnings, and imposing usage restrictions through executive orders. In fact, the U.S. Congress banned the use of the Chinese DeepSeek model within federal and defense networks last year, citing privacy and security concerns.
However, there are also significant voices opposing the introduction of such regulations.
David Sacks, a White House AI advisor and venture capitalist, strongly criticized the regulatory demands from big tech as a regulatory capture strategy aimed at eliminating competitors. He emphatically stated that weaponizing uncertain regulations as a competitive tool is absolutely unacceptable.
The WSJ observed that even within the Trump administration, differing opinions have delayed the effective enforcement of key policies, such as executive orders requiring pre-release reviews of new models from major AI companies.
Most U.S. companies, seeking to reduce AI development costs and maintain control, are rapidly increasing their adoption of high-performance open-source AI, including Chinese models.
Austin Carson, CEO of the AI policy nonprofit SeedAI, pointed out that it is crucial for U.S. companies and global users to use affordable and superior models that they can control themselves. Understanding the nature of open-source means that simply rejecting it will not win this competition.