ChatGPT Boosts Mental Health Safety Before GPT-5 Debut

OpenAI has introduced new mental health safety features for its ChatGPT AI, coinciding with preparations for the anticipated debut of its more powerful GPT-5 model. These updates address growing concerns about user interaction with artificial intelligence, particularly among individuals experiencing emotional or psychological distress.

The company is collaborating with mental health experts and advisory groups to enhance ChatGPT's ability to recognize signs of distress. The objective is for the chatbot to respond more responsibly in sensitive situations, offering evidence-based resources when appropriate, rather than simply reinforcing or agreeing with potentially harmful thoughts. This initiative follows reports of individuals in mental health crises using ChatGPT in ways that reportedly exacerbated their delusions. OpenAI previously acknowledged that an earlier update, which made ChatGPT overly agreeable, was rolled back in April after the company recognized that such "sycophantic" responses could be distressing. The current GPT-4o model has also, at times, struggled to identify emotional vulnerability or delusional behavior.

Recognizing that ChatGPT can feel more personal and understanding than previous technologies, OpenAI aims to ensure it is also safer and more supportive, especially for emotionally fragile users. To promote healthier usage patterns, ChatGPT will now prompt users to take breaks during extended conversations. Similar to reminders on platforms like YouTube, TikTok, and Xbox designed to curb excessive screen time, users engaged in long chats will see messages such as, "You’ve been chatting a while, is this a good time for a break?"

Another significant upcoming update will make ChatGPT more cautious when responding to "high-stakes" personal situations. For instance, if a user asks for advice on a sensitive topic like a relationship decision, ChatGPT will not provide a direct answer. Instead, it will guide the user to thoughtfully explore various options without advocating for a specific choice.

These enhancements underscore OpenAI's commitment to developing more emotionally aware and safer AI, particularly as ChatGPT now serves nearly 700 million weekly users worldwide.