OpenAI to launch Parental Controls and Route Sensitive Chats to safer AI models

OpenAI is rolling out new safety features to ChatGPT over the next few months, specifically to safeguard teenagers and to better assist individuals who are in distress. The company claims that it is collaborating with mental health practitioners and doctors to fine-tune the AI reaction during conversations that are sensitive. OpenAI will roll out a number of improvements to ChatGPT in the next 120 days to ensure it becomes more responsible, trustworthy, and helpful during hard times.

Rerouting sensitive conversations, including those that indicate signs of emotional or mental distress, to more complex reasoning models that are more sensitive to delicate situations is one of the largest changes. Meanwhile, OpenAI will also provide parental settings that will allow the family more control over how teens will communicate with the AI. The updates are in the larger context of the company ensuring that its tools are safer and more friendly to people of all ages.

OpenAI has established an Expert Council on Well-Being and AI, which includes experts in the fields of youth development, mental health, and human-computer interaction. This council will inform the way AI can fulfil well-being or contribute to the formulation of future protections. In tandem to this, there is a Global Physician Network of over 250 doctors representing 60 countries who are consulted on how ChatGPT should act in health-related situations. Their feedback is a direct influence on safety research, model training, and emergency intervention strategies.

The reasoning models used in OpenAI, including GPT-5 thinking, are made to take more time to analyse the context before replying. They are conditioned to the deliberative alignment and are more adherent to safety rules and less susceptible to adversarial cues. In the near future, high-risk conversations will automatically be diverted to those reasoning models so that users in distress are given more thoughtful and attentive support.

Understanding that teens are one of the first generations of AI natives, OpenAI is launching Parental Controls in the next month. Parents can connect an account with their teens, apply age-related behavioural policies, switch off such features as memory or chat history and get notified when the system notices indicators of acute distress. The tools should strike a balance between safety and trust and enable families to establish healthy guidelines on the use of AI.