The White House announced on August 1 that it had completed a voluntary safety review framework for frontier AI models, meeting the deadline set by President Trump's June 2 executive order. The framework was developed with input from Anthropic, OpenAI, and Google, which jointly reviewed and submitted modifications to the draft in late July. However, the document has not yet been made public.
The framework requires AI developers to conduct voluntary safety testing of their models before deployment, focusing on risks including misuse for biological weapons development, cybersecurity threats, and autonomous system risks. The voluntary nature of the framework reflects the tension between promoting AI innovation and managing potential risks.
The contrast with China's approach to AI governance is striking. While the US focuses on safety review of frontier models developed by a small number of companies, China is simultaneously pursuing an open-source AI strategy that has made its models the most widely used globally. Chinese companies including DeepSeek, ByteDance, and Alibaba have been releasing new models at an unprecedented pace, with five major model releases in just eight weeks.
At the 2026 World AI Conference in Shanghai, China announced the establishment of the World AI Cooperation Organization (WAICO), headquartered in Shanghai, to promote inclusive AI development and bridge the AI divide between developed and developing nations. This initiative has received support from Global South countries who see it as an alternative to Western-dominated AI governance frameworks.
The divergent approaches — US voluntary safety reviews versus Chinese open-source expansion and global cooperation — represent fundamentally different philosophies about how to balance AI innovation, safety, and global equity. The outcome of this divergence will likely shape the global AI landscape for years to come, with significant implications for technology access, economic competitiveness, and geopolitical influence.
Industry observers note that the US framework's focus on a small number of frontier model developers may miss the broader trend of AI democratization through open-source models, where Chinese companies are currently leading.
