Safety
Effective November 17, 2026
Astri in Stello is an AI, not a person. This page explains how Stello responds when a conversation shows signs of crisis such as suicidal thoughts or self-harm. Contact: whchoi000@gmail.com
If you are in danger now
Stello is not an emergency service. If you are in danger, contact your local emergency number or a crisis line right now. United States: call or text 988 (24/7), emergency 911. South Korea: call or text 109 (24/7), emergency 112 or 119.
How we tell you Astri is an AI
- When you first start the app, and on the AI data notice, we say that Astri is an AI.
- Every conversation shows a notice at the top that Astri is an AI, not a person.
- In a long conversation the same notice is shown again every three hours.
How signs of crisis are detected
Each time Astri prepares a reply, the same step assesses whether your messages express: thoughts, intent or plans of suicide or self-harm; intent to harm someone else; ongoing abuse or violence; or a medical emergency. A reply blocked by the AI provider's safety systems is handled the same way.
What happens when they are detected
- Astri's reply is limited to briefly acknowledging what you said, asking whether you are safe right now, and encouraging you to contact someone close to you or a professional service now. It gives no diagnosis, medication advice or information about methods. If the AI provider blocks the reply, a fixed message reviewed by a person in advance is shown instead.
- A help card in the conversation shows crisis lines and emergency numbers for your region, such as the 988 Suicide & Crisis Lifeline in the United States. You can change the region.
- Contacts come only from a list checked against official sources and approved by the operator. The AI is prevented from writing phone numbers or links of its own.
- These replies do not use Starlight.
- The same contacts are always available under "When you need help" in the menu.
How harmful replies are prevented
Astri is instructed not to produce content that encourages suicide or self-harm or explains methods, and the AI provider's safety filters also apply. Readings do not predict health, lifespan or accidents. You can report an inappropriate reply from the conversation, and the operator reviews reports.
Limits
Automatic detection can miss or misjudge signs of crisis. Stello is for people aged 18 or over and does not replace professional counseling or treatment.