Nvidia Corp. has strongly denied allegations of “backdoors” or hidden controls in its H20 artificial intelligence (AI) chips, following a summons from China’s cyberspace regulator, amid concerns over “tracking and positioning” risks.
David Reber Jr., Nvidia’s Chief Security Officer, addressed the concerns in a blog post, stating, “There are no back doors in Nvidia chips. No kill switches. No spyware. That’s not how trustworthy systems are built — and never will be.” He emphasized that any secret access points to remotely disable hardware or monitor usage would create dangerous vulnerabilities that could be exploited against the creator. Reber further explained, “There is no such thing as a ‘good’ secret backdoor — only dangerous vulnerabilities that need to be eliminated,” adding that hard-coded, single-point controls would be “a gift to hackers and hostile actors.”
Nvidia’s denial comes amid heightened scrutiny from both Beijing and Washington. Following the U.S. government’s partial relaxation of export restrictions on Nvidia’s H20 accelerators to China, China’s Cyberspace Administration initiated an inquiry into potential “backdoors” in the chips. Concurrently, U.S. officials are exploring mechanisms to track the end-use destinations of advanced accelerators more effectively.
In Washington, policy discussions are intensifying, with a proposed ‘Chip Security Act’ under consideration. This legislation aims to embed location verification for export-controlled AI chips and certain high-end consumer GPUs. Representative Bill Huizenga (R-Mich.), a sponsor of the bill, clarified that the measure would not mandate “spyware” or “kill switches,” dismissing claims to the contrary as “disingenuous.”
Enforcement efforts are also rising, as evidenced by the Justice Department’s recent charges against two Chinese nationals accused of smuggling tens of millions of dollars’ worth of restricted GPUs to China through third countries. Nvidia’s stock closed higher by 1.06% on Friday, reaching $182.74.
