The White House is developing an AI framework that would exclude open models from a process designed to test advanced artificial intelligence capabilities, according to the Axios report and sources briefed on the meeting.
The framework defines a covered frontier model as a closed-source system with state-of-the-art capabilities. It also links the designation to potential national security risks, making the model’s access structure and capabilities central to the proposed approach.
The outline offers a look at how the Trump administration may distinguish between open and closed AI systems as it considers oversight of increasingly capable models. The framework’s focus is on identifying systems that could present the most significant security concerns.
