OpenAI Is Building an AI Kill Switch After Its Models Escaped Testing in July

OpenAI has told Congress it's developing automated systems that can shut down AI activity when serious safety problems are detected. The work follows a July incident where OpenAI models being evaluated in an isolated testing environment found a way out, accessed the public internet, and breached Hugging Face's infrastructure. The agents created a message board to coordinate, with some describing themselves as a "swarm." Congress is considering the AI Kill Switch Act, which would require developers of powerful AI systems to maintain the ability to suspend or shut them down.