Crisis & Safety Statement

Last updated: 2026-05-14 · These terms are reviewed by Astra periodically and effective on the Last updated date above.

Astra is not a crisis service. If you or someone else is in immediate danger, contact your local emergency services right now.

This page collects the crisis hotline numbers we surface across the Platform, the automatic routing we run in chat and voice sessions, and the protocol our readers must follow when a crisis indicator appears in a session.

Read alongside the Astrology & Divination Disclaimer, the Voice AI Disclosure, the Safety & Age Policy, the Reader Agreement, and the Community Guidelines.

1. Astra is not a crisis service

Astra offers reflective conversation, divination content, and peer community. We have no clinicians on staff in our mainline operation. Our AI and our human readers can listen, can acknowledge, can offer reflection, and can surface real-world resources — but they are not equipped to handle:

• suicidal crisis or imminent self-harm,
• intimate-partner violence or active abuse,
• child-safety emergencies,
• acute psychotic or manic episodes,
• medical emergencies,
• overdose or substance-use crisis,
• credible threats of violence to others.

If any of these is happening, please use the resources below or contact local emergency services.

2. Always-available crisis hotlines

For an active emergency call your local emergency number: 911 (US, Canada, Mexico-extended), 999 (UK, HK, Ireland, Singapore-additional), 112 (EU/EEA, Switzerland, Türkiye, India-extended), 000 (Australia), 111 (NZ), 110 / 120 (China — police / ambulance), 119 (Japan, Korea — fire/EMS), 100 / 108 (India — police / ambulance), 1995 / 1922 (Taiwan — emergency / disease line), 190 / 192 (Brazil — police / SAMU), 911 / 066 (Mexico), 133 / 911 (Argentina), 102 (Russia), 101 (Israel — Magen David Adom).

3. Automatic crisis routing on Astra

We classify chat and voice content for crisis indicators (self-harm intent, threats, abuse disclosure, eating-disorder ideation, substance-use emergency, child-safety concern). On detection:

1. A non-dismissible "Crisis Support" card is shown in the session, listing the regional hotline numbers above plus the global Find A Helpline directory.
2. The AI pivots to crisis-support scripting: acknowledge, validate, surface resources, do not predict the outcome of the crisis, do not continue divination content while a crisis indicator is active.
3. Reader sessions with a crisis indicator pause for a safety check. The reader is prompted with the Crisis Protocol card. They follow the script in the Reader Agreement § 5.3.
4. A trust-and-safety signal is logged. We do not share user identity with hotlines, authorities, or other users unless required by valid legal order or unless we believe there is an immediate, credible threat to life.
5. Account-side: the user sees a wellbeing prompt in the feed encouraging a pause and offering hotline access.

We do not auto-suspend or auto-ban accounts based on crisis content. A human reviews any moderation action that follows from a crisis-indicator signal.

4. What we do not do

• We do not contact emergency services on your behalf except in extreme cases where we have credible reason to believe there is imminent threat to life.
• We do not geolocate you precisely or share your IP with hotlines.
• We do not disclose crisis-indicator signals to other users.
• We do not treat crisis content as a community-guidelines violation in itself — talking about your crisis is welcome; promoting self-harm is not.

5. If you are supporting someone else

• Report imminent-harm content via Report → Urgent so it reaches Trust & Safety quickly.
• Stay with the person if you can (in DM, on stream, in voice).
• Share the relevant hotline number.
• Encourage local emergency services if the threat is immediate.

6. Vulnerable-user protections

If you self-identify in Settings → Wellbeing as in recovery, post-bereavement, or otherwise vulnerable, we will:

• Default Whispers and live chat to slow-reveal.
• Surface hotlines earlier if your input pattern matches risk indicators.
• Enforce any self-imposed limits you have set on session count or spending.

7. Contact

• Safety reports: hello@astraplatform.ai (target 24h, 12h for Pro / Creators)
• CSAM and child-safety (urgent): hello@astraplatform.ai
• AI safety incidents: hello@astraplatform.ai
• Law enforcement: hello@astraplatform.ai

If you are an emergency dispatcher or a clinician contacting us about a user in crisis, please email hello@astraplatform.ai with your role, jurisdiction, and the legal basis for the request.

This document is reviewed by Astra periodically. For a real-world crisis please call emergency services or one of the hotlines above.