OpenAI said Tuesday it plans to route sensitive conversations to reasoning models like GPT-5 and roll out parental controls within the next month – part of an ongoing response to recent safety incidents involving ChatGPT failing to detect mental distress.