U.S. to Temporarily Exclude Chinese Open-Weight AI Models from New Safety Framework

According to Beating Monitor, the White House informed companies such as OpenAI, Google, and Anthropic during a closed-door meeting on August 4 that open-weight models will not be included in the new AI safety review framework. As a result, Chinese models like Kimi K3 and DeepSeek are not required to submit to U.S. government testing prior to release. At least for now, the U.S. will not use this framework to ban them. After Kimi K3 gained popularity, there was intense debate within the U.S. over whether to restrict Chinese open-weight models. Some officials considered placing them on trade blacklists, issuing security warnings, or even issuing executive orders specifically targeting open models. Anthropic CEO Dario Amodei also advocated that both open and closed-source models should undergo mandatory safety reviews. Subsequently, NVIDIA, alongside Microsoft, Meta, OpenAI, Google, and others, issued a public letter urging the government not to impose blanket restrictions on open-weight models. The letter argued that open weights reduce costs, enhance competition, and enable companies to deploy models on their own devices and servers. The outcome is now clear: at least in this round of safety review negotiations, the open-source side has prevailed. However, this does not mean Chinese models have received permanent exemption; the U.S. may still impose separate restrictions on specific companies regarding chips, procurement, trade, or intellectual property.