OpenAI launches Control Room with a powerful 35-second kill switch for rogue agents

OpenAI has deployed a centralized control system called Control Room to stop potentially risky AI model workloads across its research infrastructure, with the system designed to halt further model output within 35 seconds of a decision to intervene.

The system is intended to strengthen safeguards as increasingly capable AI models introduce greater safety and security risks. It connects research workloads, including agent-based model training, evaluations, and data-processing jobs, to a central service that controls whether they can continue generating outputs.

Unlike conventional monitoring systems that primarily flag suspicious behavior, Control Room can actively prevent a research workload from obtaining additional model outputs. It works alongside OpenAI’s existing systems for monitoring potentially misaligned model behavior.

The company said Control Room is already deployed across its research infrastructure, where the need for intervention is greatest. Its external model-serving infrastructure is more centralized and already has operational mechanisms to take down model deployments when necessary.

How shutdown works

Control Room connects three main components: the research workload, the inference engine, and the central Control Room service.