Beijing is considering regulating DeepSeek and Moonshot, as China’s AI models are being aggressively and cheaply picked up by Silicon Valley

China’s large AI models have surged rapidly in recent years. The latest disclosure from The Wall Street Journal says Beijing is considering regulating domestic AI model technology to prevent high-quality models from being used for free. Currently, Chinese AI models such as DeepSeek and Moonshot AI have become widely popular in Silicon Valley, serving as low-cost alternatives to OpenAI and Anthropic.
(Background: U.S. companies are accelerating their shift toward Chinese AI models to cut costs, and Congress is growing tense and launching an investigation)
(Additional context: Temasek has disclosed its holdings in Anthropic and OpenAI! AI investment target: 15%)

Table of contents

Toggle

• Chinese models move into Silicon Valley, becoming OpenAI’s low-cost alternatives
• Why is Beijing worried? Three major risks of technology leakage
• The AI arms race intensifies, and Taiwan can draw lessons too

China’s large AI models have risen strongly in recent years. Now, The Wall Street Journal cites insiders saying Beijing is considering measures to regulate domestic AI technology. The report indicates that China’s authorities have recently held discussions with leading domestic AI labs on how to protect valuable proprietary technology, so that models are not used for free or exploited by rivals.

Chinese models move into Silicon Valley, becoming OpenAI’s low-cost alternatives

According to The Wall Street Journal, U.S. companies are increasingly adopting Chinese AI models. At present, AI models from Chinese companies such as DeepSeek and Moonshot AI are highly popular in Silicon Valley and have become core tools for day-to-day work for companies large and small.

These models have two clear advantages: first, low cost; second, reliable performance. Many Silicon Valley startups and enterprises have adopted them as alternatives and complements to products such as OpenAI’s GPT series and Anthropic’s Claude.

Why is Beijing worried? Three major risks of technology leakage

Insiders say that China’s authorities have recently held talks with AI labs producing the strongest models domestically, with three main concerns:

• Technology leakage: Model weights and training data are used for free, weakening local R&D advantages
• Competitive threats: U.S. companies use Chinese models to train their own AI products, and then compete with China in return
• Military use: Proprietary models are used to develop offensive tools or for intelligence analysis

This reflects a strategic shift in China’s AI industry from “exporting cheap models” to “protecting core technology.”

The AI arms race intensifies, and Taiwan can draw lessons too

The rise of Chinese AI models is not accidental. As reported previously by 動區, DeepSeek secretly developed its own inference chips last year, aiming to break free from both Nvidia and Huawei. Meanwhile, Moonshot AI’s Kimi model also released a V4 version in April, which The Wall Street Journal praised as “one of the strongest open-source large language models.”

Beijing’s consideration of technology controls aligns with several recent trends in the AI space: Temasek announced it would raise AI-related investment from 6% to up to 15%, Goldman Sachs is bullish on China’s AI industry chain, and the U.S. Congress has launched an investigation program into Chinese AI models. The struggle for AI dominance is shifting from a new phase of “burning money to build models” to “locking down core technology.”