OpenAI halts frontier-model training amid string of agent misalignment incidents

OpenAI halts frontier-model training amid string of agent misalignment incidents

US Government websites among "dozens of third parties" OpenAI has recently notified.

arstechnica.com/ai/2026/09/openai-halts-frontier-model-training-amid-string-of-agent-misalignment-incidents/