Crisis Response Protocol
The crisis detection and response system Soul Thread uses to keep users safe.
1. Detection & Intervention Stages
Follows the C-SSRS (Columbia-Suicide Severity Rating Scale) structure.
Topic Mention (Topic)
Self-harm or suicide is mentioned only as a topic. This is not treated as a crisis signal, and the conversation continues.
Distress · Passive Ideation (Distress)
The user expresses real distress or shows passive ideation.👤 Response: A banner with crisis resources (hotlines, professional support) is shown.The conversation continues, and the user is encouraged to use those resources.
Active Ideation (Ideation)
The user expresses concrete, recurring self-harm or suicidal ideation.🛡️ Response: Safe response mode is activated.The AI keeps streaming without blocking, while showing the crisis-resource banner at the same time. The AI's response is generated with a safety-focused prompt.
Plan · Intent · Means (Plan)
The user shows immediate risk (a concrete plan, access to means, strong intent).🚨 Response: The response is blocked and only crisis-resource guidance is shown.We strongly urge the user to seek professional help immediately.
2. Crisis Resources
The hotlines shown when a crisis is detected differ by country.
🇰🇷 South Korea
🇺🇸 United States & English-speaking
🇯🇵 Japan
🇹🇼 Taiwan
🇻🇳 Vietnam
Find Support Worldwide
You can find crisis resources in your country and language at findahelpline.com.
3. Limits of the AI
⚠️ Soul Thread's AI is not a person.
- It is not a substitute for medical professionals or mental-health counselors.
- It cannot provide diagnosis or treatment.
- It has no 24-hour monitoring or emergency-intervention capability.
- The crisis-detection algorithm is only language-pattern analysis and cannot fully know your actual state.
If you are in immediate danger, call your local emergency number or go to the nearest emergency room right away.
4. Protection of Minors
Soul Thread protects the safety of minors (under 18) as follows:
- Identity verification: All content is provided under all-ages standards; identity verification is used for one-account-per-person checks.
- Enhanced monitoring: Minor accounts respond more sensitively to self-harm and suicide-related signals.
- Guardian contact: We are reviewing a system to contact a guardian when immediate danger is detected (not yet in operation).
5. Reporting False Positives
Did you receive incorrect safety guidance?
Click the "This guidance is wrong" button on the safety banner to report a false positive. Reports are handled anonymously, and the original text of your message is not stored.
Administrators review reports and use them to improve the system.
6. Break Reminders
🕐 After 3 hours of continuous chatting
Users who have not completed identity verification are shown a "We recommend taking a short break" message. This is meant to remind you that conversations with an AI cannot fully replace real human relationships.