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

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

To respect copyright, we link to the source rather than republishing the full text. Read the complete article on Ars Technica.