A new front has opened in the tense relationship between Washington and Beijing, with advanced AI chips at its center. China is now publicly accusing U.S. tech giants Nvidia and AMD of embedding hidden security risks in their processors. The allegations threaten to derail a recent, fragile agreement on chip exports, pulling the two companies deeper into a geopolitical struggle over technology and national security.
Beijing’s Allegations and Nvidia’s Firm Denial
The controversy ignited when China’s Cyberspace Administration demanded Nvidia prove its H2O AI processors are free of hidden backdoors or other vulnerabilities. Chinese state-run media quickly amplified these claims, suggesting the chips could be remotely disabled or used to steal sensitive data, creating a significant security threat for Chinese industries.
Pan Helin, an advisor to China’s Ministry of Industry and Information Technology, issued a stark warning. “If NVIDIA’s chips really have backdoor risks, that will become its ‘self-dug grave,’” he stated, signaling that the consequences could extend beyond China’s borders.
In response, Nvidia issued a strong and unequivocal denial. On August 5, Chief Security Officer David Reber Jr. stated, “There are no backdoors in Nvidia chips. No kill switches. No spyware.” The company maintains that building such vulnerabilities into its products would be self-destructive.
A Fragile Thaw in US Export Controls
These accusations come just one month after the Trump administration partially relaxed a two-year-old ban on exporting high-end AI chips to China. The deal allowed companies to ship less powerful processors to the Chinese market in exchange for a 15% fee paid to Washington.
That compromise now appears to be on shaky ground. Beijing’s security claims could be used to justify halting or slowing these limited imports, further escalating the tech trade war. For the U.S., AI chips are considered critical infrastructure, while for China, reliance on American technology is a major strategic weakness it has been trying to solve for years.
Backdoors are a Real, Historical Threat
The fear of hidden access in hardware is not new, and both the U.S. and China have been on both sides of the issue. A backdoor is a secret method of bypassing normal security, and a well-designed one is often indistinguishable from a simple software bug.
Several documented cases highlight why these concerns are taken so seriously:
- The NSA was found to have intercepted Cisco routers to install surveillance tools in the early 2010s, as revealed by the Snowden leaks.
- In 2024, a backdoor was discovered in MIFARE Classic keycards, which are manufactured by Shanghai Fudan Microelectronics.
- Chinese tech giant Huawei has faced persistent allegations that its equipment contains features allowing unauthorized access to mobile networks.
These past incidents have created a climate of deep distrust, where any claim of a hardware vulnerability is treated with utmost seriousness.
Why AI Chips Pose a Special Challenge
According to cybersecurity expert and Harvard lecturer Bruce Schneier, detecting a backdoor in a complex AI chip is exceptionally difficult. Unlike a traditional network breach that might create suspicious data traffic, a hardware backdoor could be just a few lines of hidden logic.
“For AI chips, the backdoor [could be] that if a certain thing happens, it just stops working,” Schneier explained. This type of “kill switch” would be invisible during standard testing and would not appear in any design schematics. You would only discover it when it was activated, by which point it would be too late.
This challenge is at the heart of the current standoff. With no definitive way to prove a negative, Nvidia’s denials struggle to gain traction against Beijing’s public campaign of doubt.
Caught Between Economics and National Security
The situation leaves Nvidia and AMD in an impossible position. On one hand, China remains a massive market. On the other, lawmakers in Washington are increasing pressure to restrict tech flow. The U.S. House Select Committee on the Chinese Communist Party estimates that China managed to acquire 140,000 AI chips in 2024 despite the ban, using smuggling and shell companies.
Committee chairman Rep. John Moolenaar has accused Beijing of exploiting “weaknesses in our export control enforcement system” to bolster its military. Meanwhile, Chinese officials are pushing their own domestic firms to design U.S. chips out of sensitive systems, even though their local AI industry still heavily depends on them.
With trust at an all-time low, both governments are digging in their heels, forcing tech companies to navigate a treacherous path between two opposing world powers.
Frequently Asked Questions
What is China accusing Nvidia of?
China’s Cyberspace Administration alleges that Nvidia’s H2O AI processors may contain hidden “backdoors” or “kill switches,” which could allow for remote shutdown or data theft, posing a national security risk.
How has Nvidia responded to the backdoor allegations?
Nvidia has strongly denied the claims. Its Chief Security Officer, David Reber Jr., issued a statement saying, “There are no backdoors in Nvidia chips. No kill switches. No spyware.”
Why are AI chips a major point of tension between the US and China?
The U.S. views advanced AI chips as critical technology for national security and economic competitiveness, leading to export controls. China sees its dependence on U.S. chips as a strategic vulnerability and is working to develop its own domestic alternatives while questioning the security of imported hardware.
Is it possible to detect a backdoor in an AI chip?
According to experts like Bruce Schneier, it is extremely difficult. A sophisticated hardware backdoor could be a small piece of hidden logic that does not show up in schematics or normal testing, making it nearly impossible to find until it is activated.
