OpenAI says its Control Room can stop a research job's model output within 35 seconds

OpenAI published a post on Control Room, dated 9 November 2026: a central system to track model workloads across its research infrastructure and stop them quickly when safety or security concerns arise.

Control Room is a central system to track model workloads across our research infrastructure and stop them quickly when safety or security concerns arise.

OpenAI says Control Room halts model output "within 35 seconds of a decision to intervene to stop or pause a job."

halts model output within 35 seconds of a decision to intervene to stop or pause a job.

OpenAI says Control Room connects every research inference workload, including agentic sampling for model training and evaluation, and data processing.

It connects every research inference workload, including agentic sampling for model training and evaluation, data processing, and more.

OpenAI says it requires every research workload that does agentic inference to be associated with an approved safety case.

We require every research workload that does agentic inference to be associated with an approved safety case.

OpenAI says that before fulfilling a request the inference engine asks Control Room whether the job is still authorized, and that if the job has been stopped Control Room denies the request and the engine returns a refusal.

Engines check with the Control Room service before fulfilling a request. If a job has been stopped, the service denies the request, preventing further output.

"We do not rely on the workload code to correctly perform a check of whether or not the job is still running."

We do not rely on the workload code to correctly perform a check of whether or not the job is still running.

OpenAI says the termination mechanism is implemented in the inference layer, because there are only a very small number of inference code paths across OpenAI research and workload code changes often.

We achieve this by implementing the termination mechanism in the inference layer.

OpenAI says the 35-second lag comes from a caching layer for inference authorization, and that it is working to drive the number down to zero.

The 35-second lag comes from a caching layer for inference authorization, to support the required volume of traffic. We are working to drive this number down to zero so that we can end jobs instantly.

OpenAI says its system "fails closed": if the Control Room service goes down, including through tampering, all inference requests will be denied until Control Room resumes normal operation.

our system “fails closed”, meaning that the Control Room integration in the engines is configured in such a way that if the Control Room service goes down, either due to a system error or because it is disabled for any other reason, including tampering by a bad actor, then all inference requests will be denied until Control Room resumes normal operation.

OpenAI says all of its frontier reinforcement learning jobs run with a monitor, another model that inspects everything the model being trained does and raises an alert if it detects a problem.

All of our frontier reinforcement learning jobs run with a monitor. A monitor is another model which inspects everything that the model-being-trained is doing, and raises an alert if it detects a problem.

OpenAI says that if the monitoring system detects a potential issue it may decide to stop the job by sending a message to Control Room with the job_id.

If the monitoring system detects a potential issue, depending on the details and severity, the monitoring system may decide to stop the job. In this case, the monitoring system sends a message to Control Room to stop the job, passing the job_id.

OpenAI says its recent DNS-related incident was caught by a monitor.

For example, our recent DNS-related incident was caught by a monitor.

OpenAI says fail-closed is rolled out now to all frontier model research workloads and will finish rolling out across all remaining research workloads over the next few days.

Fail-closed is rolled out now to all frontier model research workloads, and will finish rolling out across all remaining research workloads over the next few days.