OpenAI deploys Control Room kill switch to halt research models in 35 seconds
OpenAI has deployed a centralized safety system called Control Room that tracks research workloads across its infrastructure and can terminate model output within 35 seconds when safety or security issues arise.
OpenAI has introduced Control Room, a centralized monitoring system deployed across its research infrastructure designed to track workloads and quickly halt model outputs when safety or security risks are detected. According to an announcement from OpenAI, the mechanism can terminate or pause a job within 35 seconds of a decision to intervene.[4][2][5]
The platform links every research inference workload, encompassing data processing, evaluation, and agentic sampling used for training frontier models. As detailed by Wall St Engine, AI research tasks must register with Control Room before running, while separate AI monitors evaluate models for potentially dangerous or misaligned activity. If a serious violation is flagged, Control Room executes a shutdown directly at the inference layer to prevent individual jobs from bypassing the command.[4][3]
Control Room is built with a fail-safe architecture: if the system itself goes offline or experiences tampering, research inference requests are denied automatically. OpenAI is currently working to reduce the 35-second shutdown latency to zero, while reporting from Crypto Briefing suggests the rapid cutoff mechanism could set a benchmark for the industry and potentially bring increased regulatory scrutiny and higher overhead costs for frontier AI developers.[3][1]
Key facts
- OpenAI deployed Control Room, a centralized system to track research workloads and halt model outputs during safety or security interventions.
- Control Room can stop or pause a model workload within 35 seconds of an intervention decision.
- All AI research jobs must register with Control Room before running, covering training, evaluation, agentic sampling, and data processing.
- The shutdown operates at the inference layer to stop workloads from bypassing it.
- If Control Room is tampered with or goes offline, model inference requests are automatically denied.
- OpenAI is working to lower the 35-second shutdown window to zero.
- Crypto Briefing reported that the shutdown standard could lead to greater regulatory scrutiny and increased operational costs across AI development.
Sources · 5 sources
- CB
Crypto BriefingArticle ·
OpenAI builds a control room that can halt model output in 35 seconds OpenAI's rapid shutdown capability may set a new industry standard, prompting increased regulatory scrutiny and higher operational costs for AI developers. The post OpenAI builds a control room that can halt model output in 35 seconds appeared first on Crypto Briefing .
Open source - *B
*Walter Bloomberg@DeItaonePost on X ·
*OPENAI CONTROL ROOM CAN STOP MODEL OUTPUT IN 35 SECONDS *OPENAI SAYS CONTROL ROOM ALREADY DEPLOYED
Open source - WS
Wall St Engine@wallstenginePost on X ·
OPENAI BUILT A KILL SWITCH FOR ITS AI MODELS. OpenAI introduced "Control Room," a centralized system that can stop AI models during training and research if safety or security concerns arise. Here's how it works: - Every AI research job must be registered with Control Room before running. - Separate AI monitors watch frontier models for potentially dangerous or misaligned behavior. - If a monitor flags a serious issue, Control Room can terminate the job and block further model output within 35 seconds. - The shutdown mechanism operates directly at the inference layer, preventing individual research jobs from bypassing it. - The system is designed to fail safely. If Control Room itself goes offline or is tampered with, model inference requests are automatically denied. The system is already deployed across OpenAI's research infrastructure, with fail-safe enforcement active for all frontier model research workloads. OpenAI is now working to bring the 35-second shutdown window down to zero.
Open source - OP
OpenAI@OpenAIPost on X ·
We built Control Room, a central system to track model workloads across our research infrastructure and stop them quickly when safety or security concerns arise. It connects every research inference workload, including agentic sampling for model training and evaluation, data processing, and more. This works together with our misalignment monitoring systems, and halts model output within 35 seconds of a decision to intervene to stop or pause a job. Control Room is just one part of our holistic approach for safe frontier model training and deployment. https://t.co/tc0lBqFdS5
Open source - CO
Cointelegraph@CointelegraphPost on X ·
🚨 NOW: OpenAI launches Control Room, a central system that tracks model workloads across its research infrastructure and can halt them within 35 seconds if safety or security concerns arise. https://t.co/VPA1qQAkUr
Open source