OpenAI builds Control Room to stop risky model workloads quickly
OpenAI has built Control Room, a central system that tracks model workloads across its research infrastructure and stops them when safety or security concerns arise. The company says it connects every research inference workload and works with its misalignment monitoring systems, halting model output within 35 seconds of a decision to stop or pause a job. OpenAI says the system fails closed, so if Control Room goes down or is tampered with, all inference requests are denied until it resumes. Fail-closed is rolled out to all frontier model research workloads now, with the remaining research workloads to follow over the next few days.