The U.S. says China's AI progress is down to 'distillation.' But is it that clear cut?
- ID
- 25943
- Status
- summarized
- Published
- 18 Sep 2026, 9:00 PM
- Fetched
- 18 Sep 2026, 10:09 PM
- Provider
- CNBC Technology
- Category
- technology
- Original URL
- https://www.cnbc.com/2026/09/18/china-ai-progress-distillation-tech-download.html
- Source URL
- https://www.cnbc.com/id/19854910/device/rss/rss.html
Summary
- Score
- 5.5
- Created
- 18 Sep 2026, 10:10 PM
- Tags
- Audience
- ai_ml_learnersai_agent_userssaas_founders
What happened
U.S. companies and Washington accuse Chinese AI labs of using 'distillation'—training models on the outputs of more advanced, expensive models—to close the AI gap, calling it theft. But Cohere CEO Aidan Gomez and other analysts push back, arguing China's progress isn't solely from distillation and that restricted access to key technology has driven genuine innovation.
Why it matters
If distillation is a legitimate and widespread technique, builders using frontier model APIs should understand that their usage patterns could indirectly train competitors' models, and that open-weight model provenance matters when choosing foundations. The debate also signals that export controls on chips and tech may be accelerating rather than slowing Chinese AI development—a factor that could reshape which models and tools become available globally, including in Southeast Asia.
Discussion angle
If distillation is as powerful as claimed, what does that mean for Malaysian startups relying on U.S. frontier model APIs—are they subsidizing competitors, and should they care about model provenance when picking a foundation model?