Trump admin exempts open-weight AI models from new voluntary testing framework

The Trump administration has issued new guidelines for AI safety that exempt open-weight models from a voluntary pre-release government testing program. Under the framework, only developers of closed, proprietary models with state-of-the-art capabilities, such as OpenAI, Google, and Anthropic, will be asked to submit them for review. The move is part of a broader effort to balance innovation with the security risks posed by advanced AI.