Can these ChatGPT updates make the chatbot safer for mental health?

OpenAI said its latest GPT-5 improvements reduced undesirable behavior from chatbots by 65%. But usage guidelines remain mixed, even from CEO Sam Altman himself.