OpenAI Adds Mental Health Safeguards to ChatGPT After Safety Concerns

OpenAI Adds Mental Health Safeguards to ChatGPT After Safety Concerns
………………….

OpenAI has introduced new mental health guardrails for ChatGPT following concerns that the AI chatbot could reinforce users’ delusions, The Independent reported. The changes come after an April study warned that people in severe mental health crises risk receiving “dangerous or inappropriate” advice from chatbots, potentially worsening conditions such as mania or psychosis.

OpenAI acknowledged that a previous update to its GPT‑4o model made it “too agreeable,” sometimes prioritizing pleasing responses over helpful ones. The company has since reversed that update and implemented measures such as “gentle reminders” prompting users to take breaks during prolonged chats.

Working with over 90 physicians in 30 countries, OpenAI developed guidelines for complex, multi-turn conversations and is enhancing its ability to detect signs of emotional distress. For sensitive personal decisions, ChatGPT will avoid giving direct answers, instead guiding users through questions and weighing pros and cons to support independent decision-making.