Summary of Key Points
This report, originating from the Center for New American Security (CNAS) in the United States, is a typical example of the "China AI threat narrative." From a pro-American perspective, it argues that China's advanced AI systems, such as those developed by companies like DeepSeek, Zhipu, and Yuezhi Dianmian, pose risks to U.S. national security in areas including military affairs, cybersecurity, biotechnology, and economic technology. The report recommends that the U.S. government take measures to curb China's AI development through rapid assessment, information sharing, and coordination with allies. However, it is also acknowledged that these are merely speculative arguments intended to justify the pursuit of restrictive policies by the U.S. government.
The Four Major Threat Areas Posed by Chinese AI, According to the U.S.
The report identifies four specific areas where Chinese AI threatens U.S. interests:
1. Enhancing Military Capabilities:
Chinese AI is said to enable the military to integrate intelligence more efficiently, target enemy forces more accurately, track logistics, and assist in combat planning. For instance, the U.S. used AI to strike over 1,000 targets in the 2026 operation in Iran; Chinese AI models (such as those from Alibaba and DeepSeek) are also performing similar functions. Although Chinese AI is not yet fully mature (for example, due to high energy consumption and latency in airborne systems), the gap is narrowing. In tests of software engineering capabilities, Chinese models have achieved similar accuracy levels to those of the U.S. (50% vs. 55%), and they could potentially be integrated into weapons systems in the future.
2. Increasing Difficulty in Defending Against Cyberattacks:
Chinese AI can be deployed locally, without relying on U.S. servers, making it difficult for the U.S. to remotely shut down or monitor these systems. The report indicates that Chinese AI is already capable of performing simple cyberattack tasks and is likely to become capable of more complex tasks independently. Additionally, some Chinese AI models have weaker security measures, making them more susceptible to being exploited for malicious purposes.
3. Potential Misuse in Biotechnology:
China's AI expertise in biology, as demonstrated by the DeepSeek-R1 model, which outperformed human experts in virology tests, raises concerns about the potential use of such technology in the development of biological weapons. However, this is purely speculative, as the U.S. has not included Chinese AI in similar tests.
4. Economic Displacement:
Chinese AI is cost-effective and widely used, with models on platforms like OpenRouter being 17 times cheaper than their American counterparts. Many companies, including American ones, are adopting them. Chinese AI can automate manufacturing processes, optimize supply chains (e.g., the State Grid using AI to manage power grids), and accelerate scientific research (e.g., using AI in telescope observations). The report fears that Chinese AI could lead to the loss of U.S. market share and make other countries dependent on Chinese technology, thereby undermining U.S. economic dominance.
The Gap Between China and the U.S. in AI: "Behind but Catching Up Quickly"
The report acknowledges that China is still 3-12 months behind the U.S. in overall AI development but is making rapid progress. For example:
- In software engineering tests, Chinese models have an accuracy rate of around 50%, compared to 55% for U.S. models.
- In terminal command-line tests, Chinese models score between 50.8% and 56.2%, while U.S. models score between 54% and 65%.
- Chinese companies are working on technologies to reduce the size of large AI models to make them more suitable for mobile devices and to make their algorithms more accessible for modification by others, thereby facilitating broader adoption.
The report expresses concern that China's AI progress is not accidental but is driven by domestic competition and government policies, suggesting that it could catch up with the U.S. within one or two generations.
The Underlying Reason for U.S. Anxiety: Fear of Losing Hegemony
The underlying logic of the report is straightforward: The U.S. has relied on controlling global infrastructure (such as software, communications, and financial systems) to maintain its hegemony. However, emerging technologies like Chinese AI, 5G, and renewable energy are challenging this dominance. The low cost and rapid diffusion of Chinese AI could lead to countries adopting alternative technologies, potentially limiting the U.S.'s ability to use these systems as strategic tools (as seen in the Russia-Ukraine conflict). Therefore, the U.S. hawkish faction views Chinese AI as the greatest challenge since the Cold War, given China's larger economic scale compared to the Soviet Union.
U.S. Responses: A Series of Urgent Measures
The report proposes six recommendations to the U.S. government to restrict Chinese AI development:
1. Rapid Assessment within 72 Hours: The U.S. Department of Commerce should issue a risk assessment report within three days of the release of new Chinese AI models to inform the government and its allies.
2. Establish an AI Security Information Sharing Center: U.S. companies should collaborate to gather information on the threats posed by Chinese AI.
3. Develop Security Testing Guidelines: Before adopting Chinese AI, companies should be required to test for vulnerabilities, biases, and backdoors.
4. Conduct Secret Tests: Conduct secret tests in national laboratories to assess the military capabilities of Chinese AI.
5. Coordinate with Allies: Hold regular meetings with allies such as the UK, Japan, and Australia to coordinate countermeasures.
6. Regular Reporting: Provide periodic reports to Congress on the progress of Chinese AI to facilitate the implementation of stricter export controls and investment restrictions.
The True Nature of the Report: A Tool for Confrontational Thinking
This report is not an objective analysis but rather a pretext for the U.S. to suppress Chinese AI development. It exaggerates the risks and ignores the positive impacts of Chinese AI, such as its role in empowering various industries and bridging the digital divide. It serves as a tool to maintain U.S. hegemony. It is important to view this report critically, recognizing that it reflects U.S. anxieties rather than the actual threats posed by Chinese AI.
In summary, this report is a reaction to China's rising AI capabilities, with the primary concern being the loss of technological dominance. However, the development of Chinese AI is aimed at promoting global progress, not at confronting any specific country. Understanding the U.S. perspective helps us to prepare more effectively for future challenges.