“The safety work for the OpenAI o1 and OpenAI o1-mini models included safety evaluations, external red teaming, and Preparedness Framework evaluations.”