US should not treat China as enemy in AI, should work with it to keep world safe: Nvidia CEO

In the middle of tussle between China and the US for winning the AI race, arguably the most influential person in AI space has spoken. His view — cut it out guys! Nvidia CEO Jensen Huang says that instead of fighting like angry cats, the US and China should be working together in the AI space. This, he believes, will help everyone make AI safe for the world. In a podcast on Wednesday, Huang says that the US should not see China as an enemy in AI space.

His comments come even as concerns around rapidly evolving AI systems are increasing, including developments such as the Mythos AI model, which have raised fresh questions around cybersecurity and global cooperation. At the same time, countries like China and the US are increasingly using AI in military, as evident from the recent saga involving Anthropic and Pentagon.

Speaking on the Dwarkesh Podcast, Huang talks about the need for cooperation between the two countries to plan on the safe use of AI. “We want the United States to win, but I think having a dialogue and having a research dialogue is probably the safest thing to do,” he said. He added that setting boundaries is critical as AI grows more powerful, “It is essential that we try to both agree on what not to use the AI for.”

The comments from the Nvidia chief come at a time when Washington is actively debating stricter export controls on advanced AI chips to China. Policymakers have raised concerns that selling such hardware to China could be used to train models with cyber-offensive capabilities, potentially posing risks to national security.

Huang, however, suggested that limiting access to chips may not be as effective as some believe. He pointed out that China already has substantial computing resources, a strong base of AI researchers, and significant manufacturing capacity. In his view, treating China purely as an adversary could backfire, pushing the development of parallel ecosystems rather than ensuring safer outcomes.

It is important to note that China is a huge potential market for Nvidia in case the US relaxes its export-control on AI chips.

In the podcast, Huang also touched upon Mythos, a new AI model that everyone apparently is talking about. Mythos, developed by Anthropic, has reportedly identified high-severity vulnerabilities across operating systems and browsers. This capability has turned the AI model into, ahem, mythical. Anthropic believes that it is so powerful that it doesn’t warrant a public release at the moment.

Huang says this perceived capability of Mythos — its ability to find bugs no one else can see — is part of AI’s intended role. “With respect to finding bugs in software, of course, that’s what AI is supposed to do,” he said. “There are lots and lots of bugs.”

Huang suggested that as models grow more powerful, they should not be isolated. Instead, he emphasised the need for collaboration rather than isolation. He argued that AI should not be seen purely as a threat, noting that the same technology capable of finding vulnerabilities can also be used to fix them.