GLM-5.2 Becomes the First Real Test of US AI Export Controls — and Exposes the Gap

TechTimes·medium signal

Analysts argue the open-weight release of GLM-5.2 undercuts the June 12 US export ban on frontier models like Mythos: because the Chinese model is downloadable under MIT terms and benchmarks competitively on cybersecurity, restricting proprietary US models does little to limit comparable open capability abroad. The episode frames a structural weakness in weights-based export policy that builders and policymakers will now have to confront.