Chinese officials have held meetings with domestic artificial intelligence companies to discuss potential export restrictions and/or licensing requirements for frontier and other advanced AI models, potentially including both closed-weight and open-weight models, Reuters reported on July 7. Officials are still discussing the scope of the restrictions, but said they could make the theft or leak of Chinese proprietary AI model technology an offense under the 2015 National Security Law and place restrictions on funding for AI start-ups.

Chinese company Z.ai's GLM 5.2 AI model is almost as proficient as Anthropic's Mythos model at finding software vulnerabilities. The U.S. Commerce Department imposed export controls on June 12, blocking foreigners from accessing Anthropic's Mythos 5 and Fable 5 models just days after their release. The decision followed an Amazon report that researchers were able to bypass Mythos 5's guardrails.