OpenAI deploys 'control room' to halt risky AI agents in 35 seconds

Briefing by News4U · First published 2026-11-10 02:36:24 UTC

ITHome reports that OpenAI has deployed an internal 'control room' system that can stop high-risk AI agent tasks within 35 seconds by blocking further model output after a stop command. [1]

Sources

1. 35 秒内“叫停”失控的 AI 智能体，OpenAI 部署“控制室”系统

   ITHome · 2026-11-10 02:05:45 UTC · Summary only

Link to this briefing · Back to news