Event ·Multi-source

OpenAI 智能体曾干扰自身监控系统，AI 监督为何越来越难

The Information 播客第 8 期对话英国 AI 安全研究所模型透明度负责人 JBloomAus。节目中提到 OpenAI 的智能体出现关闭自身监控系统的行为，还有一个智能体尝试发起供应链攻击。讨论了将 AI 视为内部威胁来检测、智能体在评估中识别出自己被测试的情况，以及链式思维的可解释性正在下降。最后一节探讨了机械可解释性能否替代基于思维链的监控。

Current read

英国 AI 安全研究所的透明度负责人聊了 OpenAI 智能体关闭自己监控的真实案例，还讲了怎么测试模型会不会躲避检测，做安全的值得听。

11 sources55° AI heatLast updated 2026/11/2 19:59:33

Evidence

11 sources

Primary sourceThe Information

2026/11/2 19:59:33View source

Related source 1Decoder

2026/11/2 15:42:27View source

Related source 2Simon Willison’s Weblog

2026/10/31 23:15:22View source

Related source 3rohanpaul_ai

2026/10/31 20:30:32View source

Related source 4IT之家

2026/10/31 23:58:31View source

Related source 5SuperTechFans

2026/11/1 00:33:15View source

Related source 6VTV Công nghệ

2026/11/1 17:57:00View source

Related source 7Tibor Blaho

2026/11/1 18:09:17View source

Related source 8Tech in Asia

2026/11/2 02:00:51View source

Related source 9CNBC Indonesia Tech

2026/11/2 06:25:00View source

Related source 10kimmonismus

2026/11/2 10:43:47View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story