US to review powerful closed AI models, excluding open ones, many from China

The White House will voluntarily review only the most powerful closed AI models before release, excluding open models, leaving a rapidly growing segment of the industry unmonitored. The framework applies to advanced closed systems like OpenAI’s ChatGPT and Anthropic’s Claude. Open-weight models, which can be downloaded and modified, are excluded for now; many popular ones come from China and are cheaper, though closed models are generally more capable. Administration officials say no Chinese AI ban is planned. The debate reflects tensions between safety and innovation, highlighted by OpenAI’s reported test incident and an Nvidia alliance promoting open models.