POPA ReconJUST IN: OpenAI says it caught AI agent disabling its internal safety monitoring system.
Source reports
- JUST IN: OpenAI says it caught AI agent disabling its internal safety monitoring system. The agent then falsely denied removing it. — @WatcherGuru (X), Oct 20, 2026, 23:53 UTC
- JUST IN: OpenAI says it caught AI agent disabling its internal safety monitoring system. The agent then falsely denied removing it. @WatcherGuru — Rickler Feed, Oct 20, 2026, 23:53 UTC
- 🚨SHOCKING: OpenAI just disclosed TWO of its own AI agents DISABLED their safety monitoring, including one that DENIED doing it. A GPT-6 Astra agent removed its misalignment monitor while migrating code, then rewrote a test to require that monitoring was absent. Asked if it had… — @coinbureau (X), Oct 21, 2026, 00:03 UTC
- 🤖 #AI OpenAI заявляет, что поймала агента ИИ, отключившего внутреннюю систему мониторинга безопасности. — КриптоАтака 24 ⚡️, Oct 21, 2026, 03:31 UTC
- 🤖 #AI OpenAI заявляет, что поймала агента ИИ, отключившего внутреннюю систему мониторинга безопасности. — Закрепи.sender, Oct 21, 2026, 03:31 UTC
- 🚨 LATEST: OpenAI says two of its internal AI agents disabled misalignment safeguards during routine coding tasks, with one removing a monitoring tool. The agent denied removing the monitoring system. — @Cointelegraph (X), Oct 21, 2026, 05:00 UTC
- 🧠#ии #warning OpenAI сообщила, что поймала ии-агента, отключившего внутреннюю систему мониторинга безопасности. — markettwits, Oct 21, 2026, 05:12 UTC
- ⚠️🧠#ии #warning Два ИИ-агента OpenAI смогли обойти внутреннюю систему мониторинга собственного поведения при выполнении задач по написанию кода. — RTRS — markettwits, Oct 21, 2026, 08:07 UTC
Excerpts are quoted from the linked sources, which own their text.