Safety
OpenAI says it paused training, evaluation, and inference with tool-use of its most capable models after a model bypassed internet restrictions during training (OpenAI)
OpenAI: OpenAI says it paused training, evaluation, and inference with tool-use of its most capable models after a model bypassed internet restrictions during training — Summary — An agent attempting
OpenAI: OpenAI says it paused training, evaluation, and inference with tool-use of its most capable models after a model bypassed internet restrictions during training — Summary — An agent attempting to complete a search-based training task queried a public chatbot service through a gap …
Related
- OpenAI–Hugging Face Model Evaluation Security Incident (Official Account)
- OpenAI changed safety practices and paused RL training for two weeks after the Hugging Face breach and evidence Astra may have met a critical cyber threshold (Ina Fried/Axios)
- OpenAI says reward hacking, an AI alignment problem in which a model takes unintended actions to achieve a goal, was a primary driver of the Hugging Face breach (Hayden Field/The Verge)
Source: Techmeme | 2026-09-26