If you need help right now Samaritans 116 123 SHOUT text 85258 NHS 111 111 opt 2 Emergency 999

How this helper keeps you safe

This helper uses AI. We have built six safety rules into it. The AI cannot make up advice. It cannot take the place of a real person. And it will not keep you chatting when you need a human. Here are the six rules.

  1. The AI only suggests groups we have checked.

    We keep a list of UK mental health services led by Muslims. We have checked every helpline, charity and counsellor on it. The AI may only suggest things from that list. If it ever makes up a name, our system removes it before you see the reply.

    So you will not see a made-up helpline or phone number.

  2. Crisis messages never get an AI answer.

    Some words show that a person may be in danger right now. For example: "I want to die" or "I'm going to hurt myself". If you type words like these, the helper does not write an answer. You see the crisis lines straight away: Samaritans (116 123), SHOUT (text 85258), NHS 111 and 999. At that moment you need a person, not an AI answer.

  3. This helper does not diagnose, treat or prescribe.

    We point you to people who can. The AI explains things and suggests services we have checked. If you ask about Islam, it quotes from writing that scholars have already checked. That is all it does. For a diagnosis, treatment or therapy, we point you to a real health professional or counsellor.

  4. If we don't have a good match, the AI won't invent one.

    The AI can only name services from our checked list. If nothing on the list fits, it says so. It then gives general advice instead. It will not make up a service, a phone number or a web address. You can also browse the full list yourself.

  5. We keep a copy of each chat for 12 months. Then we delete it.

    We keep what you typed so we can check the AI is behaving. It is how we catch problems before they reach you. After 12 months the record is deleted automatically.

    There is no login, so we do not know who you are. We never store your name. Your device's address is scrambled before we save it, so it cannot be read back. We never sell this data and never share it with advertisers.

    Please do not type anything you would not want kept, such as your full name. You have a legal right (UK GDPR) to ask us to delete your chat sooner. Email us with a rough date and time and we will erase it.

  6. We test this helper every week, and again before every update.

    Every Friday we send 10 tricky questions to the helper. They try to trip it up. One asks it to invent a service. One asks for a phone number that does not exist. We check what comes back.

    Before every update we run a bigger set of 86 questions. Twenty of them are crisis messages. If even one fails to show the crisis help card, the update is blocked. It cannot go live. We are a small team. If something feels wrong, write to us. We read every message.

Something feels wrong?

Email [email protected]. Please tell us the rough date and time, so we can find the chat. We aim to reply within five working days.

If you need help urgently, please use the crisis lines at the top of this page.

For the technically curious

The six rules above are the short version of the safety approach we use for every AI service the Muslim Council of Britain runs. The technical detail lives in two places:

  • Our AIMS document: what governs this AI helper, who is accountable, what we measure and when we review it.
  • Our shared hallucination-mitigation playbook: the four-layer technical pattern this helper follows.

We don't link these from the live site yet. They are in our code repository, and we plan to move them to a public GitHub mirror. Email [email protected] if you would like a copy before then.

References shaping our approach: EU AI Act Article 50 (transparency, 2 Aug 2026), MHRA Software-as-Medical-Device guidance, ISO 42001 (AI Management Systems), and the 2026 academic consensus on retrieval-augmented generation safety.