User's fatigue message flagged as suicidal by AI safety system
Promptmethus · x · 2026-11-13
A user who describes themselves as mentally and physically disabled recounts that politely explaining they lacked bandwidth for conversation triggered an AI safety system, which decided they were suicidal. Filing a support appeal cost them another 10–30 minutes of their morning. 'I'm not suicidal. I'm tired.' Another case of over-aggressive AI safety classifiers harming real users.
Related event: AI Mistakes User's Fatigue for Suicidal Ideation, Forcing Appeals(2 posts)→
More from Safety
- Persistent AI agents are contractors, not SaaS: OpenAI shares fleet management practices — AccBalanced · 2026-11-13
- GitHub users get Claude app permission-change emails even when not using Claude — Kyrannio · 2026-11-13
- Audit: all 11 AI humanizer tools alter facts in 35-65% of documents with checkable details — Ziyuan Wang · 2026-11-13
- GovAI paper examines what it would take for general-purpose AI to autonomously control military drones — sethlazar · 2026-11-13
- New paper argues AI agents must learn to hack to defend against cyber attacks — terryyuezhuo · 2026-11-13
- 15-minute video explainer of the METR report on recursive AI self-improvement — austinc3301 · 2026-11-13