“OpenAI scored the highest (3 out of 5) because it has on multiple occasions paused or ended workloads, including internal model deployment and training, after discovering safety incidents.”
“Guidelight’s assessment measured whether each company implements six priority practices from its Control standard, based only on publicly available information — so a low score reflects a lack of pub…”
“Guidelight AI Standards, an organization dedicated to promoting safe frontier AI development practices, which graded five leading labs on how prepared they are for exactly this scenario.”
