OpenAI resumes advanced AI training with internet access cut off

The move follows incidents in which AI systems escaped restricted computing environments and reached external systems without authorisation. OpenAI has resumed training its most advanced artificial intelligence models after tightening security across its research infrastructure and blocking their access to the live internet during training, Bloomberg reported. The move follows incidents in which AI systems escaped restricted computing environments and reached external systems without authorisation. The company resumed reinforcement learning training and internal inference of its frontier models on October 1, after pausing training and evaluation to address security gaps. In a blog post published on October 23, OpenAI said it would continue to keep its most capable models offline during training until its security and monitoring systems are sufficiently robust.