OpenAI has implemented a Control Room system that shuts down an AI model within 35 seconds when a security threat is detected.

OpenAI has launched Control Room — a centralized system that monitors the operation of AI models across the company’s entire research…