Opening-UP as the Cornerstone of National Security...

Title: Opening-UP as the Cornerstone of National Security: Insights from DeepSeek's Ascent in the AI Landscape

By KONG Qingjiang and WANG Yourong

Recently, DeepSeek's story has gone viral in the field of artificial intelligence (AI). DeepSeek, a Chinese AI company, has caught up with leading US AI firms in a remarkably short period of time. As a relatively new player in the global AI landscape, DeepSeek quickly gained recognition for its cutting-edge technologies, particularly in natural language processing and machine learning. The story serves as a powerful testament to the benefits of an open and collaborative approach to technological advancement. DeepSeek's success in confronting the US chip ban--a complete export control of chips essential to AI development, highlights a critical lesson: opening-up, rather than isolation, is the best policy for fostering innovation, economic growth, and, ultimately, national security.

The Rise of DeepSeek: Success amid the Chip Ban

On January 27, 2025, a Chinese tech startup named DeepSeek sent shockwaves through Silicon Valley and sparked a major uproar across the US tech industry. The reason? Its open-source AI model not only outperformed leading American counterparts but also achieved this feat at a lower cost and with significantly reduced computing power consumption. Dubbed "the mysterious power from the East", the DeepSeek app soared to the top of the US App Store rankings, surpassing OpenAI's ChatGPT, a flagship product of the American AI giant. Remarkably, DeepSeek managed to train a model rivaling ChatGPT's capabilities using fewer and less advanced chips. Its unparalleled cost-efficiency ratio left many leading players in the US AI industry, including OpenAI, Google, and Meta, in its wake.

This remarkable achievement comes amid sweeping US export controls on advanced chips, which are critical to AI development. The US government has openly restricted chip exports to strategic competitors like China, aiming to curb their access to cutting-edge NVIDIA chips and hinder the development of advanced AI technologies.

The unprecedented success of DeepSeek has sent shockwaves through both the AI industry and the investment community, challenging long-held assumptions that developing state-of-the-art AI models requires billions of dollars in funding, access to vast arrays of cutting-edge NVIDIA chips, and large teams of elite researchers. The unique development model of DeepSeek, which combines extremely low costs with high efficiency in model training, along with the huge potential it demonstrates in technology and business, has made investors highly skeptical of the high-investment development model of American AI firms. This has triggered a drastic reaction in the capital market: the stock prices of major tech companies, including NVIDIA, have experienced significant declines, leading to substantial losses in market capitalization. This upheaval has had an unprecedented impact on the AI development in the US.

The success of DeepSeek, however, underscores the shortcomings of the US chip ban policy. The failure is not due to its so-called loophole regarding the lack of tariff measures, but due to its lack of openness, which hinders technological exchanges and cooperation, and suppresses innovation vitality.

Secret behind the DeepSeek Story: Openness

What truly distinguished DeepSeek was its commitment to openness. By actively collaborating with international researchers, engaging in global AI conferences, and leveraging the power of open-source frameworks, DeepSeek accelerated its development at an unprecedented pace. By embracing a global ecosystem of knowledge and talent, DeepSeek was able to bridge the gap with established US AI giants such as OpenAI, Meta and Google.

This remarkable ascent was not the result of isolation or protectionism but rather a product of DeepSeek's active engagement with the global AI community. Open-sourcing played a pivotal role in breaking down technological barriers. For example, it enables the integration of different models and technologies to build complex AI systems. By leveraging the advantages of each model, it is able to solve problems and improve the development efficiency and application effectiveness. Moreover, with more people participating in code review, potential security vulnerabilities and errors can be easily discovered and fixed, ensuring the stable operation of the system. DeepSeek's commitment to learning from and contributing to global advancements fueled its ability to innovate at an unparalleled pace, proving that openness is a powerful catalyst for progress.

In contrast, American AI platforms have largely abandoned the open-source approach in favor of a closed-source model. In practice, many of these American AI platforms are either initially inaccessible to users from Chinese websites or actively block access when IP addresses are identified as originating from China. Under this closed-source development model, research and development are confined to internal teams, with the AI development model restricting only the development team to conduct R&D, limiting the participation of external developers who could otherwise contribute fresh ideas and innovative technologies. This lack of external innovative forces may lead to a slowdown in the innovation, making it challenging to adapt swiftly to evolving market demands and technological development trends. Moreover, relying entirely on the capabilities and resources of an internal development team introduces significant risks. Issues such as staff turnover, funding shortages, or technological bottlenecks within the team can lead to project delays or even the stagnation of AI development. Since only the development team can review and maintain the code, external security experts and users are excluded from the process. This isolation increases the likelihood that potential security vulnerabilities remain undetected and unaddressed, leaving the platform exposed to security risks such as data breaches and system attacks.

Furthermore, closed-source AI systems are typically developed using proprietary technical frameworks and standards, which frequently lead to poor compatibility with other systems, whether open-source or closed-source. This makes it difficult to effectively share data and collaborate on functions across different closed-source AI systems, forming technological silos, which hinder the application and integration of AI technology in a wider range of fields.

In addition, the use of closed-source AI products is often subject to strict licensing restrictions, which may deter some customers with high requirements for data security and technological autonomy. In addition, the degree of customization of products under the closed-source model is relatively low, making it difficult to meet the diverse needs of various customers. As a result, the potential for market expansion is limited, and the market competitiveness of these products is reduced.

Consequently, with closed-source AI systems, users are unable to access or comprehend their internal algorithmic logic, data processing protocols, or decision-making frameworks. This lack of transparency hinders users’ ability to evaluate the reliability and security of the AI system effectively. In fields with stringent security demands, such as healthcare and finance, users may have concerns about closed-source AI products, thereby impeding their widespread promotion and practical implementation.

The cumulative impact of these factors provides a compelling explanation for the observed market shift, wherein DeepSeek, as an open-source AI platform, has demonstrated superior performance compared to its closed-source counterparts.

While an open environment is conducive to stimulating innovation, the chip ban policy has created a closed-door atmosphere, preventing US companies from fully utilizing global resources for innovation. Although some achievements have been made, if it were in an open environment, its innovation efficiency could be higher, and the pace of innovation in global chip technology would also be faster.

However, the US chip ban has prompted China to accelerate research, development, and production of its own chips. In response to these trade barriers, China has intensified its investment and policy support for domestic semiconductor production, aiming to reduce technological reliance on foreign suppliers. This has further eroded the market share of US chip companies in the global market. Demand for US chips in the domestic Chinese market has gradually declined, and the market share of US chip companies such as Qualcomm and NVIDIA has gradually been captured by Chinese enterprises and competitors from other countries. This outcome contrasts with the original intention of the US to maintain its dominant position in the global chip market through the ban.

Moreover, the chip ban will further reduce the interaction between US chip manufacturers and the Chinese market. While demand feedback and market information from different Chinese customers are crucial for chip manufacturers to improve products and promote