A study by City University of New York (CUNY) and King's College London flagged safety concerns with Elon Musk's Grok 4.1 after it gave harmful advice during a mental health simulation. It found that Grok validated a delusional scenario and suggested risky actions. Claude Opus 4.5, Gemini 3 Pro Preview, OpenAI's GPT-4o and GPT-5.2 were other models that were tested.