OpenAI Model Misalignment Report Von Hackernews 17. September 2026 security ORIGINAL QUELLE:openai.com Quelle: Hackernews Comments Tags: comments, ki-modell, mismanagement, openai, Security