“OpenAI is actively preventing users from accessing o1's internal reasoning traces, including sending cease and desist letters to those attempting to jailbreak the model.”