2026-10-23 · America/Los_Angeles · 官方发布 · #5

Securing our frontier model training environment

OpenAI explains how we strengthened containment for frontier model training and evaluation, including offline environments and high-compute red-teaming.

热度 47.5 / 100；排名与评分保留该期记录。