Congress Links China’s DeepSeek to Espionage, AI Theft, and Surveillance Infrastructure: A National Security Threat

In a groundbreaking report released last week, the House of Representatives’ Select Committee on China has labeled DeepSeek, a Chinese AI company, as a significant national security threat. This bipartisan investigation highlights the company’s alarming ties to the Chinese government, its illicit acquisition of technology, and its role in an expanding surveillance infrastructure. DeepSeek’s rapid rise, despite limited resources, has raised concerns about its use of stolen model architectures and illegal chip acquisitions. The committee’s findings suggest that the company’s generative AI model, which rivals the capabilities of top U.S. firms, is powered by a network designed to siphon critical data and influence global AI development in favor of China.

the Findings:

DeepSeek, founded by Liang Wenfeng, emerged with a groundbreaking AI model that stunned the tech community due to its similarity in performance to models from American giants. However, the company’s rise wasn’t fueled by innovation alone—it reportedly benefited from illicit practices, including the acquisition of stolen American AI chipsets and reverse-engineering techniques. These practices have raised red flags in both national security and the ethical development of artificial intelligence.

The House Select Committee on China’s report notes that DeepSeek operates within a Chinese-government-controlled ecosystem, with close ties to state institutions and major government-affiliated organizations like Zhejiang Lab. DeepSeek’s operations reportedly involve the collection of sensitive user data, which is allegedly routed through state-controlled telecom provider China Mobile, raising further concerns about the scope of data manipulation and surveillance.

The report also claims that DeepSeek gained access to thousands of Nvidia chips, likely in violation of U.S. export restrictions. Furthermore, testimony from OpenAI suggests that DeepSeek has employed controversial “reinforcement learning” methods to replicate and accelerate model development. This method reportedly involves bypassing guardrails in OpenAI’s models, extracting information for unauthorized use in training its own models, a process known as model “distillation.”

As a result, the committee has urged Nvidia to disclose the identities of customers in Asia who have made large-scale purchases of AI chips in recent years, as DeepSeek has reportedly purchased tens of thousands of such processors. The findings represent a stark warning about the risks posed by DeepSeek’s unchecked rise and the potential for espionage and intellectual property theft.

What Undercode Say:

DeepSeek’s rapid ascent in the AI race, coupled with the revelations about its ties to Chinese government entities, paints a concerning picture of how state-sponsored actors can leverage technological advancements for geopolitical influence. In the world of AI, where models are the backbone of future innovation, these actions could shift the balance of power, enabling China to exert disproportionate influence over global AI standards, data, and ethical practices.

The allegations about DeepSeek’s data practices are particularly troubling. In an age where personal information is a precious commodity, the potential for surveillance-driven data collection on a massive scale is frightening. If DeepSeek’s AI tools are indeed used to harvest user data and funnel it back to Chinese authorities, this could set a dangerous precedent, not only for AI companies but also for tech users worldwide. It’s a reminder of how AI, a tool that can improve lives and economies, can also be weaponized to serve broader political and strategic interests.

The issue of AI model theft and reverse engineering is another critical concern. As the committee notes, DeepSeek’s model appears to have benefited from American advancements in machine learning. This is not just an issue of corporate espionage but of intellectual property theft, where unauthorized access to proprietary algorithms and methods could lead to the creation of competitive yet dangerous technologies. The fact that DeepSeek allegedly used reinforcement learning to shortcut the development process is a clear signal of how quickly adversaries can replicate and scale AI systems without incurring the costs and risks associated with original research and development.

It’s also important to recognize the broader implications of these findings. The report paints a picture of a tech arms race where innovation is driven by espionage and covert tactics, which could lead to a global AI war. In such a competitive environment, countries and companies that fail to secure their intellectual property and data may find themselves at a severe disadvantage—or worse, vulnerable to exploitation.

From a regulatory standpoint, the U.S. has already taken action to restrict China’s access to advanced chip technologies, with entities like Nvidia under increasing pressure to track and report sales to Chinese companies. However, the committee’s report highlights the difficulties in controlling cross-border tech transfers and underscores the need for more robust, globally coordinated policies to secure AI and semiconductor supply chains.

Fact Checker Results:

1. The allegations of

2. The report accurately identifies

The claim of model distillation and learning from OpenAI’s models is supported by multiple sources, including OpenAI’s own concerns about potential misuse of their training methods.