Can Chinese AI companies curb cyber risks from open-weight models without losing their edge?

In a recent interview, AI pioneer Kai-Fu Lee addressed three common misconceptions about China's AI race efforts. Firstly, Lee emphasized that Chinese companies are primarily focused on making money, not solely on developing artificial general intelligence (AGI) or winning the AI race. Just like other tech giants such as Pinduoduo, Meituan, and ByteDance, these Chinese AI developers were initially doubted for their ability to achieve breakthroughs, but Lee reassured that the Chinese have the capability to find ways to make money.

Secondly, Lee debunked the notion that China's embrace of open-weight models and open-source technology is a strategic move by the government to compete with America. Instead, open-source technology has become popular due to market competition, and Chinese startups like DeepSeek have emerged with AI models that use significantly less computing power than their American counterparts.

DeepSeek, for instance, released a highly capable AI model in 2025 that was trained using fewer computing resources than its American competitors, challenging the assumption that Chinese companies need more resources than their American counterparts.

Lastly, Lee argued that U.S. export controls on the most powerful chips are not the primary reason for the shift in Chinese AI development towards less resource-intensive models. Instead, Chinese companies are more frugal and prioritize prudence and survival instincts over the pursuit of AGI. This contrasts with the American approach, which is characterized by heavy investments in computing power to attract top researchers and accelerate AI development.

Consequently, Chinese companies can develop and operate models at a fraction of the cost of their American competitors, making their products attractive to businesses seeking affordable AI services.