Why Open-Weight AI Is the New Frontier in the US-China Tech Race
US tech companies protest ban on Chinese open-weight AI models, citing safety concerns and lack of full open-source models.
Intelligence analysis by Qwen 2.5 (3B)
US tech companies are protesting a potential ban on Chinese open-weight AI models, citing safety concerns and the lack of full open-source models. The debate centers on the safety of these models and how the US should respond to the rise of Chinese AI technology.
Some tech companies in China are making their AI models free for anyone to use. But some people are worried that these models might be used to spy on people. The US is thinking about making it illegal to use these models. But some people think it's okay to use these models as long as you don't let anyone else see how they work.
Analysis
Safety of Open-Weight AI Models
Jason Corso, a professor of AI at the University of Michigan, notes that while building a backdoor is theoretically possible, he is not aware of any documented instances of such actions. He also points out that the safety of closed, proprietary AI models is also being questioned. Kyle Miller, a senior research analyst at Georgetown University’s Center for Security and Emerging Technology, argues that the safety of open-weight AI models is not a concern as long as they are run on one's own infrastructure. He states, 'If you’re using a Chinese open model on your own infrastructure, that simply cannot happen. You have full control over the model.'
Open-Weight vs. Open-Source Models
Aalok Mehta, director of the Wadhwani AI Center at the Center for Strategic and International Studies (CSIS), explains that the issue of open models is one of the trickiest AI policy issues. There are two types of open models: open-weight and fully open-source ones. Open-weight models are free for anyone to download, but they do not have access to the training data behind the model. Open-source models, by contrast, publish their training data and code. None of the Chinese models at the center of the debate, including those of DeepSeek and Kimi K3, are fully open-source, as all of them withhold their training data. This leaves room for worries that the models could be trained with bias or embedded with sleeper functions such as 'backdoors' that could give the Chinese government covert access.
The Role of Open-Weight Models in Defense
Kyle Miller argues that open-weight models are necessary tools to defend against black boxes. He states, 'When the model is open weight or open source, both bad actors and good actors can use it. Open-weight models allow defenders—so this includes the entire cybersecurity industry—to use these models freely without guardrails, to customize them and to optimize them for their defenses.'
The Issue of Distillation
Trump administration officials claim that the advances in Chinese open-weight AI models are due to distillation, accusing companies like Moonshot AI and DeepSeek of training their models by harvesting output from US AI labs. However, leading US industry figures like Nvidia CEO Jensen Huang argue that distillation is no reason to ban Chinese AI models. He states, 'Distillation, learning from AI, learning from other sources of knowledge, is fundamental to intelligence.'
Key points
- US tech companies are protesting a potential ban on Chinese open-weight AI models
- Open-weight models are free for anyone to use, but they do not have access to the training data behind the model
- Open-source models publish their training data and code, while open-weight models do not
- Trump administration officials claim that the advances in Chinese open-weight AI models are due to distillation
- Leading US industry figures argue that distillation is no reason to ban Chinese AI models
Open-weight AI models can help protect against bad actors and improve cybersecurity. They allow defenders to customize and optimize these models for their defenses.
If the US bans open-weight AI models, it could hurt innovation and competition in the AI space. It could also lead to a lack of transparency and accountability in the development of these models.
