Nvidia CEO Jensen Huang has cautioned that China already possesses the computing power and data center infrastructure needed to train an AI model on par with Anthropic’s Claude Mythos, raising potential concerns for global cybersecurity.
Speaking on the Dwarkesh Patel podcast on Wednesday, Huang addressed whether China’s access to advanced chips could enable it to develop a model with cyberoffensive capabilities similar to Mythos — a scenario that could pose risks to US national security.
He added that Mythos itself was trained using what he described as “fairly mundane” computing resources.
“The amount of capacity and the type of compute it was trained on is abundantly available in China, so you just have to first realize that chips exist in China.”
Anthropic restricted access to its latest AI model in April after it uncovered thousands of software vulnerabilities across major operating systems and browsers, sparking concerns about how such capabilities could be exploited in cyberattacks. A comparable model developed in China could pose significant risks if misused.
Jensen Huang emphasized that China’s available computing power is “enormous,” underscoring the scale of resources that could be directed toward developing similar AI systems.
“They have datacenters that are sitting completely empty, fully powered. You know, they have ghost cities, they have ghost datacenters too. They have so much infrastructure capacity. If they wanted to, they [could] just gang up more chips.”
A call for dialogue, not conflict
Huang added that China produces around 60% of the world’s mainstream chips, boasts a deep pool of top-tier computer scientists, accounts for roughly half of global AI researchers, and has ample energy resources to support large-scale computing.
“Victimizing them, turning them into an enemy, likely isn’t the best answer,” he said, while acknowledging, “They are an adversary.”
“We want the United States to win. But I think having a dialogue and having research dialogue is probably the safest thing to do.”
On Tuesday, US Treasury Secretary Scott Bessent praised Mythos as a major breakthrough that could help maintain America’s lead over China in the AI race. “This Anthropic Mythos model was a step function change in abilities, learning capabilities,” he said, according to Bloomberg.
Claude Mythos poses a real threat
Anthropic released its findings on Claude Mythos Preview on April 7, raising concerns about the model’s potential misuse in cyberattacks due to its ability to identify and possibly exploit zero-day vulnerabilities. The company noted that 99% of the vulnerabilities uncovered by the model remain unpatched.
Further evaluation by the AI Security Institute (AISI) on April 13 found that Mythos could “execute multi-stage attacks on vulnerable networks and discover and exploit vulnerabilities autonomously” — tasks that would typically take human experts days to complete.
AI-driven attacks powered by models like Mythos could also pose serious risks to financial institutions, particularly banks that rely on decades-old legacy systems, Reuters reported Tuesday.
Separately, Anthropic disclosed in November last year that a “Chinese state-sponsored group” had attempted to misuse its Claude Code tool to target around 30 global entities, successfully breaching a small number of them.
