OpenAI releases ChatGPT safety updates for harm detection
OpenAI released safety updates to ChatGPT that improve detection of suicide, self-harm, and harm to others in multi-turn dialogues and separate sessions. The changes enable the model to identify subtle or evolving risk cues and generate short-lived safety summaries to guide responses. The updates build on more than two years of collaboration with mental health experts and the existing safe completion framework, allowing refusal, de-escalation, or redirection in high-risk cases while preserving normal handling for routine interactions.
#127Boaz Barak@BOAZBARAKTCSGetting mental health right is about complexity rather than simplicity. It is about understanding the context, and collaborating with experts, and careful measurements. Really proud of @declangrabbmd and the team on their work, that makes a meaningful difference.
Declan Grabb, MD@declangrabbmdsharing out new work that helps ChatGPT better recognize context in sensitive conversations and respond safely in these complex/nuanced scenarios-- both within long conversations and across separate conversations! see blog post for details: https://openai.com/index/chatgpt-recognize-context-in-sensitive-conversations/