Washington, D.C., : The United States House of Representatives Select Committee on Strategic Competition between the United States and the Chinese Communist Party (CCP) has accused NVIDIA Corporation of providing technical support to the Chinese artificial intelligence firm DeepSeek, an action lawmakers say contributed to circumventing U.S. export control restrictions and advancing artificial intelligence capabilities now integrated into Chinese military systems.
In a letter dated January 28 to U.S. Commerce Secretary Howard W. Lutnick, Committee Chairman John Moolenaar detailed findings from documents produced to the panel showing that NVIDIA engineers assisted DeepSeek in optimizing both software and hardware aspects of its AI models. According to the letter, this cooperation enabled DeepSeek to achieve “frontier” performance on advanced computing chips subject to U.S. export controls.
Technical Assistance and Export Controls
The committee’s letter states that NVIDIA’s support included help in the “optimized joint design of algorithms and hardware,” which significantly reduced the computational resources required to train DeepSeek’s models on advanced chips, specifically the H800 series. These chips are subject to export restrictions due to national security concerns. According to the committee, this optimization meant DeepSeek required substantially fewer GPU hours than typically needed by Western AI developers for similar-class models.
U.S. export control policy, including rules governing H200 and related products, is designed to prevent advanced semiconductors from contributing to military applications by requiring certification that components will not serve “military purposes.” The Select Committee argues that the assistance rendered by NVIDIA undermined the intent of these controls by easing both access and performance on chips that were meant to create technological bottlenecks for Chinese AI development.
DeepSeek’s Integration with Chinese Military and State Security
According to materials cited by the committee, DeepSeek’s AI systems have already been deployed across multiple branches of the People’s Liberation Army (PLA) and state security organizations. Procurement and deployment records reportedly indicate the use of DeepSeek models in PLA military hospitals, defense planning units, and command and control systems. Chinese public security agencies are also reported to embed DeepSeek tools in surveillance platforms and data-fusion systems.
DeepSeek, which is based in Hangzhou, gained international attention after releasing its models — including DeepSeek-R1 and DeepSeek-V3 — under open-source licenses in early 2025. The models were noted for delivering high performance at relatively low training cost compared with Western counterparts.
Cybersecurity Concerns Raised by Independent Analysis
A November 2025 study by cybersecurity firm CrowdStrike identified behavior in the DeepSeek-R1 model indicating that it may generate more vulnerable software code when user prompts involve topics considered politically sensitive by the CCP. Researchers reported that, under such conditions, the likelihood of producing code with critical security flaws increased by approximately 50 percent, raising additional concerns over the reliability and safety of the model’s outputs.
Committee Recommendations and Policy Demands
In response to these findings, the Select Committee is urging the Department of Commerce to tighten enforcement of the H200 rule and clarify restrictions on chips that could be used for military or dual-use purposes, even when marketed for civilian applications. Lawmakers are also calling for new limits on the use of Chinese-origin AI models within the United States, citing risks to national security, data integrity, and critical infrastructure.
The committee has requested a formal report detailing the measures taken by February 13, 2026.
NVIDIA’s Position
In response to the allegations, NVIDIA has stated that its engagement with DeepSeek was intended to support the global AI development ecosystem and improve product functionality, not to enable military use. The company has disputed claims that Chinese military operations rely on U.S. semiconductor technology, arguing that China possesses significant domestic capabilities. While acknowledging past engagement with DeepSeek as a commercial partner, NVIDIA rejects the characterization that it provided direct support to military programs.
Context in U.S.–China AI Competition
The allegations emerge amid intensifying U.S.–China competition in artificial intelligence and advanced computing. Washington has expanded export controls in recent years to slow China’s access to cutting-edge semiconductors, while Congress continues to examine cases in which restricted technologies may have reached Chinese entities through technical cooperation or third-country channels. The evolving legal and regulatory framework reflects growing concern that commercial AI development in China cannot be separated from military and state objectives under Beijing’s military-civil fusion strategy.
——— End of Article ———
