The White House informed major AI developers that open-weight models, such as Meta's Llama and Nvidia's Nemotron, will be exempt from its new voluntary AI safety testing framework. Closed models developed by companies like OpenAI, Google, and Anthropic will remain subject to review.