Crisis Prevention & Response Protocol

Last updated February 28, 2026

relif is committed to supporting users who may be experiencing emotional distress. This document describes how our AI companion Liam detects crisis indicators, what happens when they are detected, and where to find immediate help.

If You Need Help Right Now

If you or someone you know is in immediate danger, please contact emergency services (911 in the US, 112 in Europe, or your local emergency number).

• US/Canada: Call or text 988 (Suicide & Crisis Lifeline)
• Crisis Text Line: Text HELLO to 741741 (US), 686868 (Canada), or 85258 (UK)
• International: Visit findahelpline.com to find a helpline near you

relif is NOT an emergency service. We cannot contact emergency services on your behalf. We do not monitor conversations in real-time.

Table of Contents

1. How Liam Detects Crisis Indicators
2. Our Response Protocol
3. Crisis Hotline Directory
4. For Parents & Caregivers
5. Our Limitations
6. Data Handling During Crisis Events
7. Regulatory Compliance

How Liam Detects Crisis Indicators

Liam, our AI wellness companion, includes a crisis detection system designed to identify language that may indicate a user is in distress. Here is how it works:

What We Look For

Our system analyzes conversation messages for patterns that may suggest:

• Expressions of suicidal ideation or self-harm
• Severe emotional distress or hopelessness
• Language indicating immediate danger to self or others

Multi-Language Detection

Crisis detection operates across all 8 languages supported by relif:

• English (en)
• Hebrew (he)
• Arabic (ar)
• German (de)
• Spanish (es)
• French (fr)
• Portuguese (pt)
• Russian (ru)

Each language uses culturally appropriate keyword patterns and phrases, accounting for the different ways distress is expressed across cultures.

How Detection Works

1. Pattern matching: Messages are checked against language-specific crisis keyword databases
2. Context awareness: Liam considers the broader conversation context, not just isolated words
3. Immediate response: When crisis indicators are detected, the response is triggered within the same conversation turn — there is no delay

What We Do NOT Do

• We do not record, listen to, or monitor audio or video
• We do not have real-time human moderators reading conversations
• We do not contact law enforcement, emergency services, or third parties based on detection
• We do not use crisis detection to restrict, punish, or flag user accounts

Our Response Protocol

When Liam detects crisis indicators in a conversation, the following steps occur automatically:

Step 1: Acknowledge & Validate

Liam responds with empathetic, validating language. The goal is to make the user feel heard, not judged.

Step 2: Present Crisis Resources

A dedicated crisis support widget appears in the chat, prominently displaying:

• Local crisis hotline number (based on the user's language/region)
• Direct call button — one tap to dial the hotline
• Link to findahelpline.com — for users in countries without a listed hotline
• Crisis Text Line information where available

Step 3: Grounding Support

Liam offers grounding techniques (such as guided breathing) to help the user through the immediate moment, while emphasizing that professional support is available and recommended.

Step 4: Continued Availability

Liam remains available for the rest of the conversation. Crisis detection does not end the conversation or lock the user out. The crisis support resources remain accessible.

Crisis Hotline Directory

Below are crisis hotlines for countries where relif is available. All numbers are free and confidential.

North America

Europe

Middle East

Asia

Oceania

South America

Africa

Not Listed?

Visit findahelpline.com to find a crisis helpline in your country.

For Parents & Caregivers

If your child or someone in your care uses relif, here is what you should know:

• Age requirement: relif requires users to be 13 years or older. Users under 13 are blocked during onboarding.
• Minor protections: Users aged 13–17 receive additional safeguards, including session break reminders after extended use and enhanced AI disclaimers.
• Liam is AI: Liam is an artificial intelligence wellness companion, not a therapist, counselor, or medical professional. Liam's responses are generated by AI and may be inaccurate.
• Crisis detection: If your child expresses distress in a conversation with Liam, crisis hotline resources will be shown automatically.
• No real-time monitoring: We do not have human staff monitoring conversations. If you are concerned about your child's safety, please speak with them directly and consult a mental health professional.
• Data access: You can request access to or deletion of your child's data by contacting support@myrelif.com.

Our Limitations

It is important to understand what relif can and cannot do:

What relif IS

• A wellness companion app designed to help users manage anxiety
• An AI-powered tool that offers breathing exercises, journaling, mood tracking, and conversational support
• A resource that connects users to professional crisis services when distress is detected

What relif IS NOT

• Not an emergency service — we cannot dispatch help or contact 911/112
• Not a medical device — relif does not diagnose, treat, cure, or prevent any medical condition
• Not a replacement for therapy — relif does not provide therapy, counseling, or clinical treatment
• Not a real-time monitoring service — conversations are not reviewed by human staff in real-time
• Not available offline — crisis detection requires an internet connection

Important

If you are in immediate danger or experiencing a medical emergency, call your local emergency number immediately. Do not rely on relif or any app for emergency assistance.

Data Handling During Crisis Events

When a crisis detection event occurs, we record the following for compliance and safety improvement purposes:

• Timestamp of the event
• Language in which the crisis was detected
• Country/region (based on locale, not GPS)
• Action taken (e.g., hotline displayed, call button tapped, resources dismissed)

We do not store:

• The specific message content that triggered detection
• Audio, video, or screen recordings
• Personal identifying information in the audit log

Crisis audit data is stored securely and is accessible only to authorized administrators. This data is used for aggregate reporting and system improvement, never for individual user profiling.

Regulatory Compliance

This protocol is published in compliance with:

• California SB 243 (effective January 1, 2026) — requires AI platforms used for mental health support to maintain and publish a crisis detection and response protocol, and to refer at-risk users to crisis services.
• FTC Act Section 5 — truthful representation of app capabilities and limitations.

relif will submit annual reports to the California Office of Suicide Prevention as required by SB 243, beginning July 2027. These reports will contain aggregate, anonymized data about crisis detection events — never individual user information.

Contact Us

If you have questions about this crisis protocol or suggestions for improvement:

• Email: support@myrelif.com
• Website: myrelif.com

Anthropic Cloud LLC · 2810 N Church St, Suite 74948, Wilmington, DE 19802-4447 · relif is a registered trademark.