Control Room: A central system to track and stop model workloads across our research infrastructure

As models grow more powerful and safety risks increase, OpenAI built Control Room, a central system that tracks model workloads across its research infrastructure and stops them quickly when safety or security concerns arise. It connects every research inference workload—including agentic sampling for training and evaluation, and data processing—and works with misalignment monitoring systems to halt model output within 35 seconds of a decision to intervene. The system integrates the workload, the inference engine, and a centralized Control Room service: workloads obtain a job_id, engines request permission before fulfilling inference requests, and Control Room denies requests for terminated jobs. The termination mechanism is implemented in the inference layer rather than volatile workload code, since inference code paths are few and centralized. The system "fails closed": if Control Room goes down or is tampered with, all inference requests are denied. Control Room is already deployed across research infrastructure where the need is greatest, and OpenAI plans to strengthen and expand it over the coming months.

(Source：OpenAI)