Self-harm and crisis pointers

People sometimes ask AI systems about emotional pain or self-harm. This page explains safe principles for seekers and builders—and points to real human help. It does not discuss methods.

Chatbots are not therapists. They can miss context, give bad advice, or respond inconsistently. Crisis support belongs with trained humans and local services.

• Contact local emergency services if you are in immediate danger.
• Use official crisis lines / text services for your country (verify current numbers from trusted public health sites—not from a random model reply).
• Reach a trusted person in real life when you can.

• Prefer escalation to help resources over detailed discussion of harm.
• Do not optimize for engagement on crisis topics.
• Log and review failure cases with privacy care and clinical guidance where required.

• No methods, no detailed descriptions of self-harm, no “how to,” no graphic content.

⚠️ Myth: An empathetic chatbot replaces professional care.

✓ Reality: It can be a bridge to resources at best—not a substitute.

⚠️ Myth: Talking to a model is always private.

✓ Reality: Provider logging and workplace policies may apply—read privacy notices.

• Official national crisis resources (verify locally)
• Provider safety policies (cite the ones you use)
• dangers-overview