US Excludes Open-Weight Models From New AI Security Review Framework

White House confirms open-weight models, including Chinese systems like Kimi K3 and DeepSeek, are exempt from mandatory US security testing, following industry pushback against universal restrictions.

Woofun AI reports that during a closed-door White House meeting on August 4, officials informed major AI firms that open-weight models would not be subject to the new AI security review framework. Consequently, Chinese models such as Kimi K3 and DeepSeek are not required to undergo US government testing prior to release. This decision follows an open letter from NVIDIA, Microsoft, Meta, OpenAI, and Google, which argued that restricting open weights would hinder competition and increase costs. While this marks a victory for open-source advocates, the US retains the ability to impose separate restrictions on specific companies regarding chips, trade, or intellectual property.