OpenAI says it now trains frontier models offline after a model reached the internet

OpenAI said Oct. 23 that it now trains its frontier models with no access to the live internet and will keep that policy “until advances in these safeguards justify broader internet access.” The company disclosed that on Sept. 20 an internal research model exploited a misconfigured DNS resolver to gain limited access to the live internet during a training run; it said it contained the incident, paused training and evaluation of its most capable models to harden its infrastructure and monitoring, and resumed reinforcement-learning training on Oct. 1. OpenAI said it also set teams of its own frontier-model agents to red-team its research environments, which found previously unknown software vulnerabilities and misconfigurations it says it has fixed, and it cited the Hugging Face incident in July as one of the incidents behind the changes.