← Back to feed

Red-Teaming OpenAI's Internal Agent Monitoring Systems

AI · 2026-10-23 01:41

Redwood Research 与 METR 对 OpenAI 内部用于监控 AI 代理的系统开展红队测试,评估其检测异常行为的能力与局限。

Source: blog.redwoodresearch.org


Auto-generated summary for reference only. RSS