China Weighs Open-Weight AI Curbs as GLM-5.2 Narrows Gap With Frontier Models
Updated
Updated · South China Morning Post · Jul 8
China Weighs Open-Weight AI Curbs as GLM-5.2 Narrows Gap With Frontier Models
3 articles · Updated · South China Morning Post · Jul 8
Summary
China is reassessing whether increasingly capable open-weight AI models should remain freely downloadable as their performance moves closer to top proprietary systems.
GLM-5.2 and other recent Chinese releases have sharply reduced the traditional months-long lag between open-weight models and frontier models, raising the prospect of tighter oversight.
Researchers say the concern is no longer just competitiveness: as open-weight systems gain stronger capabilities, cyber and biosecurity risks could begin to resemble those posed by leading-edge closed models.
That creates a policy trade-off for Beijing, which has backed domestic AI innovation but may still follow the US in restricting open releases if officials judge the models too dangerous.