Chinese open-source models are temporarily excluded from the U.S. AI safety review framework Led by Jensen Huang, NVIDIA, Microsoft, Meta, OpenAI, and Google jointly petitioned and successfully convinced the White House: Chinese open-source models are temporarily excluded from the AI safety review framework. Kimi K3 and DeepSeek dodged a bullet—they don’t need to be tested by the U.S. government before release. The open-source camp won this round of the battle, but chip and trade restrictions may still come separately.