Nvidia AI Chip Export Trial & Kill Switch Rejection
Two Chinese nationals are currently facing trial in California, accused of illegally exporting hundreds of Nvidia H100 AI chips to China without the necessary licenses. This legal action unfolds as Nvidia, a leading manufacturer of these critical components, simultaneously pushes back against growing political pressure to embed “kill switches” or tracking features into its AI hardware.
According to the U.S. Department of Justice, Chuan Geng and Shiwei Yang allegedly orchestrated a scheme to send millions of dollars’ worth of Nvidia AI chips to China between October 2022 and July 2025, deliberately circumventing American export controls. These chips are foundational to modern AI infrastructure, serving as the backbone for training advanced large language models. Reports indicate that some shipments were routed through transit hubs in Singapore and Malaysia in an apparent attempt to obscure their ultimate destination.
The individuals are said to have operated a company named ALX Solutions, which was reportedly established shortly after the U.S. implemented new export restrictions in 2022. Over two dozen shipments were officially dispatched to freight forwarders in Southeast Asia, yet payments for these transactions originated directly from companies in China and Hong Kong, suggesting a concerted effort to conceal the chips’ final recipients. In one notable instance, a shipment valued at $28.4 million was declared for Singapore, but U.S. officials were unable to locate the purported recipient. Both Nvidia and Super Micro Computer, the server manufacturer that allegedly supplied ALX with more than 200 H100 chips, have stated their strict adherence to export regulations and their cooperation with ongoing investigations. Nvidia emphasized that it only engages with verified partners and that any diverted products receive no company support, updates, or services.
Amidst these legal proceedings, a broader debate is unfolding concerning the security and control of AI hardware. Politicians and policymakers have increasingly advocated for the inclusion of “kill switches” or tracking capabilities in AI chips to prevent their misuse. However, Nvidia has unequivocally rejected such proposals, outlining its objections in a recent blog post.
David Reber Jr., Nvidia’s Chief Security Officer, contends that integrating these features would introduce severe security risks. He argues that kill switches and tracking functions could become attractive targets for malicious actors, thereby creating significant vulnerabilities within critical AI systems. Instead, Nvidia champions established cybersecurity principles such as “defense in depth,” which involves layering multiple protective measures rather than relying on a single point of centralized control. The company highlights that its chips are deployed in highly sensitive sectors, including healthcare, aviation, and autonomous vehicles, where a centrally controlled kill switch could pose catastrophic risks.
Reber draws a parallel to the U.S. government’s ill-fated Clipper chip initiative of the 1990s, a project that aimed to embed government access into encrypted communications but was ultimately abandoned due to critical security flaws. Nvidia also distinguishes its chips from consumer technologies like “Find my phone” features, stressing that the latter are optional software tools, not permanent hardware controls. The company asserts its reliance on proven security concepts and transparent software tools for monitoring, diagnostics, and patch management, all of which remain under user control. Nvidia maintains a firm stance against any form of hardware backdoor.
The ongoing trial and Nvidia’s principled opposition to hardware kill switches underscore the complex interplay between national security concerns, technological innovation, and corporate policy in the rapidly evolving landscape of artificial intelligence.