OpenAI@OpenAISep 29, 2026, 14:07How we think about securing frontier RL training runs: https://t.co/yrvjpfa7Pd打开原帖#511482