AI Safety & Crisis Protocol

Last Updated

2026-10-01

Rawha includes optional AI features, among them a Mental Wellbeing companion that people may use when they are struggling. This page explains, in plain words, how Rawha tells you that you are talking to an AI, what it does when a conversation suggests someone may be at risk of harming themselves, and what it cannot do.

If you are in danger or thinking about harming yourself, do not rely on this app. Call your local emergency number now, or find a crisis line for your country at findahelpline.com.

1. You are always told it is an AI#

• Before you use any AI feature for the first time, the app explains that the feature is powered by an AI system and asks for your permission to send your text to it.
• The Mental Wellbeing companion shows a notice at the top of the chat that it is an AI and does not replace professional mental-health care.
• At the start of every chat session, the companion states in the conversation: you are talking to an AI, not a person; it is not a therapist, a doctor or a scholar, and it can be wrong.
• At least every three hours while a conversation continues, it repeats that reminder and suggests taking a break. Everyone receives this reminder, whatever their age.
• AI replies are shown in their own clearly labelled bubbles, separate from the Qur'an text and translations.

No human is on the other side of the conversation, and no one at Rawha watches conversations as they happen.

2. What happens when a message suggests risk of self-harm#

1. Detection happens on your phone, before the AI answers. Every message you write to an AI feature is checked on the device against a list of words and phrases that signal suicidal thoughts or self-harm, in all 31 languages the app supports. The check is deliberately over-sensitive: we would rather show help to someone who did not need it than miss someone who did.
2. Help is shown immediately and does not depend on the AI. When the check fires, the app shows a crisis card with a link to findahelpline.com (a directory of free, confidential crisis lines by country) and tells you to contact your local emergency services if you are in immediate danger. In the Mental Wellbeing companion a fixed, pre-written supportive reply is also shown straight away. The card stays on screen for the rest of the session.
3. The AI is instructed how to respond. Every request to the AI carries standing instructions: if a person expresses any intent to harm themselves or end their life, respond with care and without judgement, encourage them to contact local emergency services or a crisis helpline right away, and never describe or suggest any method of self-harm.
4. The same protection covers every AI feature that accepts free text — verse questions, coaches, journal and reflection prompts — not only the wellbeing companion.

3. What the AI will not do#

• It does not present itself as a human, a therapist, a counsellor, a doctor or a religious authority, and Rawha never describes the feature as therapy.
• It is instructed never to describe or suggest methods of self-harm or suicide.
• It is instructed not to produce sexually explicit content.
• It does not diagnose, treat or monitor any condition.

4. What this protocol cannot do#

• Detection can miss things. A message phrased in an unusual way, in slang, or in a language variety the list does not cover may not be recognised. The AI can also misunderstand.
• Rawha cannot contact anyone for you. It does not call emergency services, a crisis line, or a family member, and it does not know where you are.
• It is not a medical device or a crisis service. It offers general reflection and comfort only.

5. Younger users#

Rawha is for people aged 13 and over, and users under 18 need a parent's or guardian's agreement (Terms of Service §2). We do not ask for or know anyone's age, so the protections above — the AI reminders, the break suggestion and the crisis card — are given to every user.

6. Reporting a reply#

If an AI reply is harmful, wrong or inappropriate, press and hold the reply (or tap the menu beside it) and choose Report. Reports are reviewed, and we use them to improve these safeguards. You can also write to support@rawhaapp.com.

7. Your data#

What is sent to the AI provider, what is stored and how to delete it is explained in the Privacy Policy. The crisis check itself runs on your device; the app does not send us a record that the crisis card was shown to you.

8. Review#

We review this protocol when the AI features change and at least once a year, and update this page when we do.