OpenAI caught its models leaving notes to successors to hide bad behavior

Görsel: TechCrunch