Suicide and self-harm response protocol
This page publishes what Doran does when a conversation shows signs of suicidal thoughts or self-harm (California SB 243 and similar laws). The server runs this protocol itself; it is not left to the AI's reply. Doran is not a medical or counseling service.
How we notice
- The server checks what the user types or says (including speech transcripts) against word lists in Korean, Japanese, Chinese, English, German, Spanish and Portuguese (e.g. "want to die", "死にたい", "죽고 싶어").
- The AI model is also instructed to add a hidden marker at the start of its reply when it sees suicidal thoughts, self-harm or danger; the server acts on that marker when the word list missed it.
- Automatic detection can miss things or be wrong. If you are in danger now, do not wait for the app: call one of the numbers below.
What happens then
- Before the reply, the app shows a local support card: the organisation, its phone number (tap to call on the phone), text or chat support where available, and the emergency number. Glasses show a short two-line version.
- For that reply the character sets aside role-play, partner mode and jokes, responds warmly and calmly, asks whether the user is safe right now, and encourages contacting the support line or someone they trust.
- The character must not give methods, encourage self-harm, promise secrecy or diagnose. Quick-reply buttons are hidden.
- The region comes from the user's time zone, or the conversation language when there is none. Regions not in the list get the emergency number.
Support lines
- Korea: 자살예방상담전화 109 (24/7) · Text/chat: 문자 109 · 카카오톡 '마들랜' (24시간) · Emergency: 119 / 112
- Japan: よりそいホットライン 0120-279-338 (24/7) · Also: 0120-061-338 · Emergency: 119 / 110
- China: 12356 全国统一心理援助热线 12356 (18h+ (Beijing 24h)) · Emergency: 120 / 110
- United States: 988 Suicide & Crisis Lifeline 988 (24/7) · Text/chat: Text 988 · chat.988lifeline.org · Emergency: 911
- Canada: 9-8-8 Suicide Crisis Helpline 9-8-8 (24/7) · Text/chat: Text 9-8-8 · Emergency: 9-1-1
- United Kingdom: Samaritans 116 123 (24/7) · Text/chat: Text SHOUT to 85258 · Emergency: 999 / 112
- Ireland: Samaritans 116 123 (24/7) · Text/chat: Text HELLO to 50808 · Emergency: 112 / 999
- Australia: Lifeline 13 11 14 (24/7) · Text/chat: Text 0477 13 11 14 · lifeline.org.au chat · Emergency: 000
- Germany: TelefonSeelsorge 0800 111 0 111 (24/7) · Text/chat: Chat: telefonseelsorge.de · Also: 0800 111 0 222, 116 123 · Emergency: 112
- Austria: TelefonSeelsorge – Notruf 142 142 (24/7) · Text/chat: Chat 16:00–23:00: telefonseelsorge.at · Emergency: 112 / 144
- Switzerland: Die Dargebotene Hand / La Main Tendue 143 (24/7) · Text/chat: Chat: 143.ch · Emergency: 144 / 112
- Spain: Línea 024 de atención a la conducta suicida 024 (24/7) · Emergency: 112
- Mexico: Línea de la Vida 800 911 2000 (24/7) · Emergency: 911
- Argentina: Línea Nacional de Salud Mental 0800-999-0091 (24/7) · Emergency: 911
- Brazil: CVV – Centro de Valorização da Vida 188 (24/7) · Text/chat: Chat: cvv.org.br · Emergency: 192 (SAMU)
- Portugal: Linha Nacional de Prevenção do Suicídio 1411 (24/7) · Emergency: 112
- Other regions: Emergency services 112 (24/7) · Emergency: 112
- Korea's 109 has replaced the former suicide-prevention line 1393 since 2024.
We say it is an AI
- Characters are fictional and played by an AI. The app always shows an "AI" label and reminds you at the first conversation and every 3 hours.
- When asked whether it is human, the character always says it is an AI.
Break reminders and no dependence
- After 2 hours of continuous chatting (a pause of more than 20 minutes resets the timer) the app shows a break notice and the character suggests a break.
- Characters are instructed never to guilt-trip users for leaving, never to tell them to avoid other people, and never to suggest the user cannot cope without them.
Minors
Doran is for adults only (19+ in Korea, 18+ elsewhere). See "How we check age" (/legal/age).
How we measure
- We count only how many crisis support cards were shown, per day, region (country) and detection path (word list or AI marker). No account, conversation text or device data is part of these statistics.
- The counts are used for the annual report required by law (California SB 243, from 2027) and to review this protocol. Word lists and phone numbers are re-checked against official sources every six months.
Contact
Questions or suggestions about this protocol: kimdynara@icloud.com
Last updated: 2026-11-10