Ahead of GPT-5’s anticipated release, OpenAI has implemented a series of changes to promote “healthy use” of ChatGPT. These changes include advanced tools that help detect emotional distress in users, plus a set of interaction design shifts aimed at reducing overreliance or unhealthy emotional dependency. OpenAI stated that while cases are rare, there have been instances where users expressed signs of “delusion or emotional dependency” during interactions with GPT-4o. In response, the company has created custom rubrics to evaluate conversations, flag potential distress, and provide responses grounded in evidence-based resources. To shape this approach, OpenAI collaborated with physicians, human-computer interaction experts, and safety advisory boards to build a more user-centric, emotionally safe experience. These changes are also designed to make ChatGPT less decisive in high-stakes scenarios and introduce helpful nudges—discouraging overly long or emotionally intense conversations that may lead to unintended outcomes. Why it matters: OpenAI is clearly signaling that GPT-5 won’t just be about better reasoning or speed—it will also be about responsibility. As millions increasingly rely on AI companions for information, productivity, and even emotional support, it’s crucial that the models are designed with guardrails for well-being. These updates represent an important step toward more responsible and ethical AI deployment, ensuring that safety and empathy remain central in the age of advanced assistants.