Krux
May 16, 2026
ChatGPT Now Remembers When You're in Crisis
Published: May 16, 2026 at 12:13 AM
Updated: May 16, 2026 at 12:13 AM
100-word summary
OpenAI rolled out safety updates that let ChatGPT track risk signals across conversations. The model now creates temporary safety summaries when it detects distress, then uses that context to respond more cautiously in later messages. Internal tests showed ChatGPT improved safe responses by 50% in suicide and self-harm scenarios during long conversations. The system focuses narrowly on serious harm: it will de-escalate, refuse dangerous details, or redirect to crisis resources. The summaries expire quickly and aren't used for personalization. It's live now in GPT-5.5 Instant. Your chatbot can finally tell the difference between a bad day and a pattern.
What happened
OpenAI rolled out safety updates that let ChatGPT track risk signals across conversations. The model now creates temporary safety summaries when it detects distress, then uses that context to respond more cautiously in later messages. Internal tests showed ChatGPT improved safe responses by 50% in suicide and self-harm scenarios during long conversations. The system focuses narrowly on serious harm: it will de-escalate, refuse dangerous details, or redirect to crisis resources. The summaries expire quickly and aren't used for personalization. It's live now in GPT-5.5 Instant.
Why it matters
Your chatbot can finally tell the difference between a bad day and a pattern.