OpenAI gives us a glimpse into new security features for ChatGPT that respond in a more helpful manner during sensitive conversations.
OpenAI is rolling out new features in ChatGPT over the next month, including parental controls and automatic activation of reasoning models for sensitive conversations. The company is collaborating with medical experts worldwide for this purpose.
Support for Mental Distress
OpenAI aims to make ChatGPT more useful in situations where users experience mental or emotional stress. Therefore, the system will soon automatically switch to reasoning models, such as GPT-5 thinking, when it detects signs of ‘acute distress’. These models are designed to take more time for context and reasoning. This should lead to more reliable answers in sensitive conversations.
read also
GPT-5 Gets a Warmer Personality and Fewer Restrictions
The models are trained using a method called ‘deliberative alignment’. This technique is intended to ensure they better handle safety guidelines and are less susceptible to inappropriate input. OpenAI will gradually activate conversations with such models and evaluate them carefully.
Parental Controls
Additionally, there will be extra attention for young people. According to OpenAI, teenagers are among the first ‘AI natives’ and often use AI independently. Therefore, the company will soon introduce parental controls in ChatGPT. Parents can link their teenager’s account to their own, adjust basic settings such as disabling memory and chat history, and receive notifications if their child is in a crisis moment.
Medical Experts
To support these new features, OpenAI is collaborating with an Expert Council on Well-Being and AI. This council advises on youth development, mental health, and human-computer interaction. Furthermore, the company receives support from a global network of more than 250 doctors from 60 countries. Among them are psychiatrists, pediatricians, and general practitioners who provide input on how the models should respond in the context of mental health.
Over the next 120 days, OpenAI plans to share further steps within this initiative. The rollout of parental control features and the use of reasoning models in crisis situations mark the first phase of that plan.