White House officials are discussing a possible executive order that could restrict government use of advanced open-weight AI models and potentially ban or delay the strongest releases.
June 9 talks included debate over whether open-source models should get capability-based exemptions, but the policy focus appears tied to fears that Chinese-origin open models are nearing frontier performance.
The report says any cutoff would likely target models above a capability range likened to GPT-5.5, Claude Opus 4.8 or GLM-5.2, with that threshold expected to be tested within 6 months.
Distillation concerns and broader model-capability fears are converging in Washington, with critics arguing proposed limits would favor closed-model companies and damage the emerging U.S. open-model ecosystem.
A U.S.-only clampdown would be hard to enforce globally, the report argues, raising the prospect of slower domestic open-model development without preventing foreign or illicit access.