번역 대기 중입니다. 완료되면 자동 업데이트됩니다.
OpenAI explains how we strengthened containment for frontier model training and evaluation, including offline environments and high-compute red-teaming.
This is a feed excerpt. Read the original source for the full report. Editorial review is pending.