“OpenAI's System Safety team focuses on improving the alignment and safety of current models by enhancing refusals, fixing jailbreaks, and improving monitoring.”