Table of contents
On July 31, 2025, the Cyberspace Administration of China (CAC) summoned NVIDIA to explain potential security vulnerabilities identified in its H20 chips, specifically designed for the Chinese market. This summons comes amid ongoing tensions between the United States and China over advanced computing technologies and artificial intelligence.
Accusations Related to "Tracking" and "Remote Shutdown" Capabilities
According to several Chinese media outlets and cybersecurity experts cited by the local press, NVIDIA's H20 chips allegedly integrate tracking and remote shutdown functions, potentially allowing unilateral device deactivation or localization. These technological capabilities had previously been mentioned by some U.S. lawmakers, who support the integration of control mechanisms in semiconductors exported to third-party countries, including China.
According to statements relayed by official media, these tracking technologies are now "mature" and operational in the latest generations of NVIDIA's AI chips, particularly the H20, designed as a restricted version of the A100/H100 chips to comply with U.S. export restrictions.
A Regulatory Response Based on Chinese Legislation
Relying on the Cybersecurity Law, the Data Security Law, and the Personal Information Protection Law, the Chinese regulator has demanded a formal response from NVIDIA regarding potential risks associated with these components. The American company must provide technical and documentary evidence relating to the alleged flaws, as well as a demonstration of compliance with local legislation on digital sovereignty and critical infrastructure security.
The July 31 meeting is part of a broader context of China's reaffirmation of technological control, which has recently tightened its requirements on foreign IT equipment used in state-owned enterprises, critical infrastructure, and strategic research projects.
NVIDIA in a Delicate Position
The company, whose products are at the heart of the global generative AI explosion, is once again at the center of geopolitical issues. Already constrained by U.S. authorities to limit the capabilities of its GPUs intended for China, NVIDIA has been trying for months to maintain its presence in this strategic market without violating Washington's export rules or Beijing's cybersecurity regulations.
The H20 model was designed to address these constraints while maintaining acceptable performance for large AI models. Accusations of vulnerabilities or "backdoors" could further complicate the company's already precarious balance between opposing regulations.
Cet article publirédactionnel est publié dans le cadre d'une collaboration commerciale