Helping ChatGPT Better Recognize Context in Sensitive Conversations
OpenAI, Thursday, May 14th, 2026
OpenAI updates ChatGPT to better recognize contextual risk signals in sensitive conversations about self-harm and suicide.
OpenAI announced safety improvements that help ChatGPT recognize when risk emerges gradually over the course of conversations by identifying subtle or evolving warning signs. The company developed "safety summaries"-short, factual notes about earlier safety-relevant context-created by a model trained for safety reasoning and used only in rare, high-risk situations.
Developed in collaboration with mental health experts including psychiatrists and psychologists, the updates improved safe-response performance by 50% in suicide and self-harm cases within single conversations and by 52% in harm-to-others cases across multiple conversations.
The improvements enable ChatGPT to distinguish between ordinary interactions and situations requiring caution, such as de-escalation or redirection toward support resources.