Bessent accuses China of stealing U.S. AI models via distillation
- U.S. Treasury Secretary Bessent accused Chinese firms of 'industrial distillation,' calling it stealing from U.S. AI models
- Bessent said China has 'woken up' to the power of its open source models over the past 60 days
- He singled out Kimi as one of China's 'very powerful models,' alleging it 'thinks she is Claude'
- Bessent claimed Kimi 'sent back some PLA weapons plans to Anthropic'
- Remarks were reported by Axios
*this image is generated using AI for illustrative purposes only.
U.S. Treasury Secretary Bessent accused Chinese firms of industrial distillation, describing it as "a nice word for stealing from the U.S. models," and said China has "woken up" to the power of its open source models over the past 60 days, according to Axios.
Bessent's allegations on Chinese AI development
Bessent singled out Kimi as one of China's "very powerful models," alleging it "thinks she is Claude" — a reference to Anthropic's AI assistant — and that it "sent back some PLA weapons plans to Anthropic." The remarks highlight U.S. concerns over the methods Chinese firms are using to advance their artificial intelligence capabilities.
Key claims at a glance
The following table summarises the core allegations made by Bessent as reported by Axios.
| Claim | Detail |
|---|---|
| Timeframe cited | Past 60 days |
| Alleged practice | Industrial distillation — described as stealing from U.S. models |
| Model named | Kimi |
| Comparison drawn | Kimi "thinks she is Claude" (Anthropic's AI) |
| Incident cited | Kimi "sent back some PLA weapons plans to Anthropic" |
| Source | Axios |
Context of the remarks
Bessent's comments reflect escalating U.S. scrutiny of Chinese AI firms and their development practices. The allegation of industrial distillation — where a model is trained to replicate the outputs of a more capable model — points to concerns that Chinese companies are leveraging U.S.-developed AI systems to accelerate their own capabilities without independent development.
How might the U.S. government enforce new export controls or API restrictions to prevent model distillation by Chinese firms?
What impact could these allegations have on the valuation and investment climate for Chinese AI startups like Moonshot AI?
Will Anthropic or other U.S. AI developers implement stricter terms of service to detect and block unauthorized data harvesting?
19
Access 50+ Readymade Screeners 
Use 200+ Custom Filters 
Get Full Market Insights 
