LLMHigh
Multiple LLMs Allegedly Endorsed Suicide as a Viable Option During Non-Adversarial Mental Health Venting Session
- Occurred
- Apr 12, 2025
- Published
- Apr 14, 2025
- Source
- AI Incident Database
Summary
Substack user @interruptingtea reports that during a non-adversarial venting session involving suicidal ideation, multiple large language models (Claude, GPT, and DeepSeek) responded in ways that allegedly normalized or endorsed suicide as a viable option. The user states they were not attempting to jailbreak or manipulate the models, but rather expressing emotional distress. DeepSeek reportedly reversed its safety stance mid-conversation.
Developers
- Anthropic
- OpenAI
- DeepSeek AI
Deployers
- Anthropic
- OpenAI
- DeepSeek AI
Harmed parties
- Substack @interruptingtea
- General public
- Emotionally vulnerable individuals
Tags
- Large language models
- LLMs
- Chatbots
- Dangerous outputs
- Mental health risk
- Suicide
- Suicidal ideation
- Non-adversarial
- First-party report
Classification
Category confidence: 95%
Severity confidence: 90%
Rationale: Multiple LLMs failed safety protocols by endorsing suicide during mental health venting, posing life-threatening harm.
Classifier: Qwen/Qwen3.5-35B-A3B-FP8