Building a More Helpful ChatGPT Experience for Everyone

OpenAI is strengthening safety and support features in ChatGPT to more effectively respond to moments when users experience mental and emotional difficulties. Over the next 120 days, the company plans to carry out intensive improvement work, focusing on helping users in crisis situations.

To this end, OpenAI operates the Expert Council on Well-Being and AI and the Global Physician Network, made up of more than 250 physicians across 60 countries. Based on advice from these experts, the findings will be reflected in research and model training so that AI can support users' well-being and adhere to safety guidelines.

In particular, reasoning models such as GPT-5-thinking and o3 are being utilized. When a real-time router detects sensitive signals in the conversation context, such as acute distress, it connects the user to a reasoning model capable of deeper thought and contextual understanding, in order to provide more helpful and safer responses.

Additionally, to protect teenagers who are the 'AI-native' generation, Parental Controls features are set to launch within a month. This will allow families to set healthy AI usage guidelines appropriate to a teenager's stage of development.