Nvidia CEO: China can match Mythos AI, urges cooperation

Jensen Huang says China can match Mythos AI, urging US-China cooperation on AI safety and open dialogue amid cybersecurity concerns.

Nvidia CEO Jensen Huang emphasized that China has the computational resources, infrastructure, and chip manufacturing capacity to develop AI models comparable to Anthropic’s Mythos. In recent interviews, Huang noted that the hardware and data center capacity used to train Mythos are readily available in China, which produces about 60% of the world’s mainstream chips and has significant unused data center capacity, including so-called “ghost datacenters.” Huang warned that China’s ability to mobilize these resources could allow it to rival or match Western AI advancements. He also pointed out that Mythos was trained on relatively ordinary compute, suggesting that the barriers to entry are not as high as some believe. Amid concerns about the cyberoffensive capabilities of Mythos, which can identify and exploit software vulnerabilities, Huang urged the US to engage in direct dialogue with China on AI safety. He stressed the importance of cooperation and communication between American and Chinese AI researchers to establish ethical boundaries and address security risks associated with advanced AI systems.