Nvidia Reassures: No Backdoors in Its Chips Amid Security Concerns

Nvidia has announced that its chips have no backdoors that could allow unauthorized access to the devices, as reported by CNBC.

In July, the company expressed its intention to resume sales of its H20 AI processors in China after receiving the necessary approval from the U.S. government.

Subsequently, Chinese regulators summoned Nvidia representatives for questioning to assess any potential risks associated with the chips.

Nvidia met with Beijing officials on July 31 to address concerns related to national security.

«Our priority is cybersecurity. Nvidia chips do not have any backdoors that would enable remote access or control by others,» stated a company representative on August 1.

China’s Cyberspace Administration (CAC) has requested that Nvidia provide clarifications and documentation regarding security threats, including possible vulnerabilities and backdoors in the H20 chips supplied to the country.

The regulator reminded that there had been previous concerns about significant vulnerabilities in the American company’s AI processors. It was also noted that U.S. lawmakers have advocated for mandatory tracking of advanced semiconductor exports.

The CAC added that American AI experts have indicated the presence of sophisticated «tracking and positioning» technologies, as well as «remote shutdown» capabilities in Nvidia’s computing chips.

In May, Republican Senator Tom Cotton and a bipartisan group of eight Congress members introduced the U.S. Chip Security Act, which stipulates that semiconductor companies like Nvidia implement security measures and geolocation verification features in their advanced AI processors.

Democrat Bill Foster, a co-sponsor of the bill in the House of Representatives, along with independent technical experts, told Reuters that the technologies for chip tracking already exist and are integrated into Nvidia’s products.

It’s worth noting that companies in China plan to acquire over 115,000 advanced Nvidia chips to power new facilities and further train AI models.