How did the government decide OpenAI’s frontier model was safe to release?

OpenAI is releasing a new advanced language model for public access, but the government approval process for how it decided the model was safe remains unclear to outside observers and experts. Multiple researchers and policy figures have stated they do not understand what the requirements are for getting government approval to release frontier models, or which agencies are responsible for evaluating them. An executive order laid out a roadmap for evaluating frontier models, but the specifics have not been filled in, and six cabinet agencies have been instructed to determine a final process by early August. The lack of transparency about safety evaluations raises concerns about whether true safety experts are playing enough of a role in the model release process, and whether decisions depend too heavily on personal connections between company leaders and administration officials.