Safety & Ethics
OpenAI deploys Control Room to track and halt research model workloads
On November 9, 2026 OpenAI described Control Room, a fail-closed system already running on its research infrastructure that tracks inference jobs and can stop model output within 35 seconds when safety or security concer…