Towards safety cases for frontier AI training

本文介绍了 OpenAI 发布的 Towards safety cases for frontier AI training 模型，涵盖技术保障、运营实践及误alignment 调查，旨在构建前沿 AI 训练的安全防线。