Huang Renxun Smiles: U.S. Does Not Currently Plan to Ban Chinese Open Source Model

According to Pertinence AI's monitoring, during a closed-door meeting at the White House on August 4, it was communicated to companies such as OpenAI, Google, and Anthropic that open-weight models would not be included in a new AI security review framework. Chinese models like Kimi K3 and DeepSeek are therefore not required to undergo testing by the U.S. government before release. At least for now, the U.S. will not ban them under this framework.

Following the popularity of Kimi K3, there was a fierce domestic debate in the U.S. about whether to restrict Chinese open-weight models. Some officials considered options such as a trade blacklist, security warnings, and even issuing executive orders specifically targeting open models. Anthropic CEO Dario Amodei also argued that both open and closed models should undergo mandatory security reviews.

Subsequently, NVIDIA, along with companies like Microsoft, Meta, OpenAI, and Google, issued an open letter urging the government not to universally restrict open models. The letter argued that open weights can lower costs, enhance competition, and enable companies to deploy models on their own devices and servers.

The outcome is now largely clear: at least in this round of security review games, the open-source advocates have won. However, this does not mean that Chinese models have received a permanent exemption. The U.S. may still take separate actions against specific companies, such as restrictions on chips, procurement, trade, or intellectual property.