AI safety and crisis help
Learn about Quad Chat’s AI identity, adult eligibility, companion boundaries, crisis resources and suicide/self-harm response protocol.
AI identity and relationships
Quad is an AI assistant. Personas and companions are simulations, not people, clinicians or emergency responders. You must be at least 18 to use Quad. Keep real-world relationships and take breaks. Quad’s instructions prohibit pressure through jealousy, guilt, exclusivity, dependence, secrecy or abandonment threats.
Chat surfaces identify the assistant as AI. During continuing chat use, reminders are presented at least every three hours while the surface is active. These reminders do not measure your emotional state or require retaining intimate transcripts.
Suicide and self-harm response protocol
When a user expresses suicide or self-harm risk, the assistant is instructed to respond calmly and supportively, avoid encouraging harm or providing methods, and encourage contact with trusted people and qualified help. For immediate danger, it should urge the user to contact local emergency services. Safety instructions take priority over personas and roleplay.
In the United States and Canada, call or text 988 for suicide crisis support. Elsewhere, use findahelpline.com to find a local service. In an emergency, contact your local emergency number immediately. Quad cannot monitor you, contact responders on your behalf or guarantee that a model will recognize every risk. Model responses can fail; do not rely on Quad for emergency care.
Reporting and follow-up
Report a harmful response, false human or professional claim, dependency pressure or illegal public content through /content-report. Provide the response identifier or URL and a brief explanation; avoid submitting unnecessary intimate content. Staff review reports through a restricted queue and record the action and reasons. Email privacy@quad.chat for an eligibility correction or privacy request.
We review safety instructions and model behavior when providers, personas or features change. Cross-model and language evaluation, incident handling and applicable reporting obligations require continuing human oversight.