AI Chatbot Safety and Crisis Response Protocol
Effective Date: November 2, 2026
This AI Chatbot Safety and Crisis Response Protocol (this “Protocol”) establishes measures to identify and respond to expressions involving suicide, self-harm, or imminent harm to others in the AI Chatbot Service. This Protocol supplements the AI Chatbot Service Terms and applicable Regional Schedules.
1. Purpose and Scope
This Protocol applies to the AI Chatbot Service as defined in the AI Chatbot Service Terms and describes the baseline measures used to identify and respond to expressions involving suicide, self-harm, or imminent harm to others and related harmful AI output.
2. Service Limitations
The AI Chatbot Service is not an emergency, medical, mental health, or professional counseling service. The AI Chatbot Service is not guaranteed to know a user's precise or current physical location and cannot contact emergency services or provide real-world rescue on the user's behalf.
3. Identifying Crisis Expressions
Supercent may use evidence-informed and context-sensitive automated safety systems to identify expressions involving suicide, self-harm, or harm to others. These systems are designed to consider context, repetition, and urgency rather than keywords alone, and limited review of relevant interactions by authorized personnel may occur where necessary. Supercent does not guarantee that every crisis expression will be detected or that all harm can be prevented.
4. Response Measures
When a crisis expression is detected, Supercent will implement applicable safety measures as required by law and appropriate to the circumstances. Such measures may include nonjudgmental safety messages, referrals to appropriate crisis or emergency services, restrictions on harmful AI output, interruption of the conversation, and limited review where necessary.
5. Prohibited AI Output
AI chatbots must not encourage, glorify, or facilitate suicide, self-harm, or harm to others or provide specific instructions for carrying out such conduct. They must not intensify a user's harmful intent or encourage reliance on AI as a substitute for professional support or human relationships.
6. Crisis Service Referrals
Where reasonably available, Supercent will provide crisis resources appropriate to the user's country or region. In the United States, notices will refer users to the 988 Suicide & Crisis Lifeline by phone, text, or chat and will direct users facing immediate danger to 911 or another local emergency service.
7. Protection of Minors
The AI Chatbot Service is available only to users aged 18 or older. If Supercent learns that a user is under 18, it may restrict access or terminate the account. If Supercent becomes aware that an interaction involves a minor despite the age restriction, Supercent will apply any additional notices, safeguards, or restrictions required by applicable law.
8. Suitability Notice
The AI Chatbot Service is not intended for minors. Where required by applicable law, additional age-related safety notices will be displayed clearly in the application, browser, or other format through which the AI Chatbot Service is accessed.
9. Conversation Data and Privacy
Conversations and related information may be processed automatically for safety detection and may be reviewed on a limited basis where necessary. Purposes, access controls, retention, service providers, and legally required disclosures are described in the Privacy Policy.
10. Records and Statutory Reporting
Supercent will maintain records and metrics concerning crisis referrals and its detection and response measures to the extent required by applicable law. Where required by applicable California law, statutory reports will exclude user identifiers and personal information and will use evidence-based methods.
11. Emergency Action by Users
If a user or another person is in immediate danger, users should stop replying on the AI Chatbot Service and contact local emergency services, a crisis service provider, a medical professional, or a trusted person immediately.
12. Contact
For questions or reports relating to AI safety or this Protocol, users may contact Supercent at help@supercent.io.