Amid an intensifying rivalry between the United States and China over AI supremacy, Beijing is considering tightening export controls on AI and semiconductor technology. China, which upended the AI model race dominated by OpenAI and Anthropic with the release of DeepSeek last year, appears to be moving to lock down the potential outflow of its technology after gaining fresh confidence from the "Kimi shock."
The Financial Times reported Tuesday that China's Commerce Ministry is in discussions with AI companies about ways to prevent Western nations from acquiring the country's leading technology firms and promising startups. The ministry is also discussing measures to restrict companies such as Alibaba, ByteDance and Zhipu from transferring core data overseas to train their models abroad, and to prevent foreign users from downloading the model weights of Chinese AI systems.
China's Commerce Ministry is also considering restrictions that would bar overseas semiconductor manufacturers — including Qualcomm and TSMC — from producing advanced chips based on designs developed by Chinese companies such as Huawei, Alibaba and ByteDance. Officials are also intensively discussing limits on foreign acquisitions of domestic companies with strategic technologies such as AI, seeking to avoid repeating what they view as a misstep: failing to block Meta's attempted acquisition of Manus in time.
The moves are seen as a follow-up to China's growing confidence in the AI race against the United States, buoyed first by the "DeepSeek shock" and now by the "Kimi shock." The discussions can also be read as an attempt to put domestic industry protections in place before the two countries hold AI talks.
According to Reuters, the United States and China have agreed to hold government-level talks on the AI industry before Xi Jinping's U.S. visit on Sept. 24. Xi and US President Donald Trump agreed to hold the AI talks at a summit in Beijing in May. Discussions are underway to finalize the arrangements, and Treasury Secretary Scott Bessent is expected to lead the US negotiating team.
The AI talks are still in early preparation. The United States is reviewing participants, venue and agenda while simultaneously keeping pressure on Chinese AI companies.
In an interview with Fox Business that day, Bessent warned that if allegations that Chinese AI models were developed by "distilling" American models are confirmed, the US could impose sanctions on Chinese companies.
"We are finding 'watermarks' of American large language models in many Chinese models. This is unacceptable," he said. "We will be looking into this in the coming days or weeks." Bessent said that if Chinese AI companies did distill US models as alleged, it could be considered "intellectual property (IP) theft," adding, "We have the ability to sanction them for this theft."
Both countries are weighing sanctions options over concerns that the other side's frontier AI models could threaten their industries or national security. The United States already restricts exports of advanced AI chips to China. Washington is also reportedly considering limiting domestic companies' use of Chinese-made AI models following the release of Kimi K3, a low-cost, high-performance AI model from Chinese startup Moonshot AI.
kate01@heraldcorp.com
