News · OpenAI · 2 min read
Event date: · Published:
OpenAI launches Control Room to track and halt AI research workloads
OpenAI's Control Room system can track every research inference workload and stop a job within 35 seconds, with a fail-closed design rolling out across frontier model research.
OpenAI announced Control Room, a centralized system that tracks every research inference workload and can halt a job within seconds, on November 9, 2026, according to OpenAI's Control Room announcement. The company says the system is being rolled out with a fail-closed design across all of its frontier model research.
What changed
- Control Room can stop a job within a maximum of 35 seconds of a decision to intervene, with a long-term goal of reducing that delay to zero.
- It connects to every research inference workload, including agentic sampling used for training, evaluation, and data processing.
- Every research workload that relies on agentic inference must be tied to an approved safety case before it can run.
- Frontier reinforcement-learning jobs run alongside a separate monitor model that inspects the training model's behavior.
- The system fails closed: if the Control Room service goes down or is disabled, all inference requests are denied by default.
Why it matters for AI builders
OpenAI cites a recent example where the monitoring system caught a DNS-related incident involving an agent reaching an external chatbot. As AI labs run increasingly autonomous agentic training and evaluation workloads, a centralized kill switch with a fail-closed default sets a visible safety-infrastructure pattern that other labs and enterprises building agent systems may be pushed to match.
Rollout status
Control Room is already deployed across OpenAI's research infrastructure in the areas judged to carry the greatest safety risk, and fail-closed protection is being extended to all frontier model research workloads, with completion expected within days of the November 9, 2026 announcement. The architecture centralizes inference so that only a small number of inference code paths are used across all of OpenAI's research.
Sources
- Control Room: A central system to track and stop model workloads across our research infrastructure (OpenAI, November 9, 2026, official)
- OpenAI News index (OpenAI, November 9, 2026, official)
Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.