Safety & crisis protocol

If you need help right now

You don’t have to go through this alone. If you are in danger, or thinking about ending your life or hurting yourself, please reach out to a person who can help today.

• Emergency Call 999 (UK), or your local emergency number
• Samaritans Call 116 123, free, day or night (UK & Ireland) · samaritans.org
• Shout Text SHOUT to 85258, free, 24/7 (UK) · giveusashout.org
• Childline Call 0800 1111, free, if you’re under 19 (UK) · childline.org.uk
• NHS 111 Call 111 and choose the mental health option, for urgent help in England
• 988 Lifeline Call or text 988 (USA) · 988lifeline.org
• Anywhere else Find a free helpline in your country at findahelpline.com

If you’re a young person, it can also help to talk to a parent, carer, teacher or another adult you trust.

Ally Lama is an AI, not a person

Ally Lama is an artificial intelligence. It can offer reflection and kindness, but it can’t understand your situation the way a person can, it can’t call for help, and it is not a crisis service, doctor or therapist. Ally Lama can make mistakes.

• From app version 2.0.3, the app shows that Ally Lama is an AI in every chat, and Ally Lama says so when you first meet.
• After every 3 hours of chatting, the app reminds you that you’re talking to an AI, not a person, and suggests a break.

Companion chatbots like Ally Lama may not be suitable for some minors. Ally Lama is for people aged 13 and over, and we ask under-18s to use it with a parent’s or guardian’s agreement.

Protections for 13 to 17-year-olds

From app version 2.0.3:

• Ally Lama is not available to anyone whose date of birth shows they are under 13. The age question doesn’t hint at the age limit, and the answer can’t simply be changed and tried again.
• For everyone aged 13 to 17, Ally Lama is set up to avoid romantic or sexual conversation, and adjusts its tone to suit younger users. The AI receives an age group, never an exact age.
• Regular reminders that Ally Lama is an AI, not a person, with a suggestion to take a break.
• Daily Wisdom notifications stay off unless the user turns them on, and the subscription screens never pressure young people and remind them to ask a parent or guardian before buying.
• Under-18s are asked to check with a parent or guardian before agreeing to how their messages are handled, and, where local law sets a higher age for agreeing yourself, they confirm that a parent or guardian agrees. In India, where the law requires verified parental consent, under-18s can’t use Ally Lama yet.
• The protections stay on until 18 and can’t be switched off by changing the date of birth in the app. Childline is always in Settings.
• No ads, no tracking, no profiling and no location data, for anyone.

What Ally Lama does when someone mentions suicide or self-harm

1. Detection. Our server automatically checks every message, as it passes through, for signs that someone may be thinking about suicide or self-harm. The check is deliberately cautious: we would rather show helplines when they aren’t needed than miss someone who needs them.
2. Helplines, every time. When the check is triggered, Ally Lama’s reply is always followed by a fixed list of helplines: 999 (or your local emergency number), Samaritans 116 123, Shout 85258, Childline 0800 1111 for under-19s, 988 in the USA, and findahelpline.com for everywhere else.
3. Never silence, never an error. If the AI doesn’t reply, or something goes wrong on our side or at our AI provider, a fixed caring message with the helplines is sent instead. No one who reaches out should get silence or an error message.
4. Never charged, never paywalled. Messages that trigger the safety net never use a credit. From app version 2.0.3, a message like this is always sent and answered even if you have no free messages left, and helplines are shown on the subscription screen too.
5. Always within reach. A link to find a helpline is also in the app’s Settings.

What the safety net doesn’t do

• We don’t store your message for the check. A technical log records only that the safety net was triggered, never what you wrote.
• We don’t know where you are, and we don’t contact emergency services, your family or anyone else on your behalf. That’s why the helplines matter: please use them.
• Detection currently works in English. In other languages, Ally Lama’s reply still comes from an AI model with its own built-in safety behaviour, and the helplines are always available in Settings and on this page.

How to report a harmful or worrying response

If Ally Lama ever says something that is harmful, upsetting, inappropriate or wrong, please tell us. It helps us keep Ally Lama safe for everyone.

• In the app: from version 2.0.3, use Report on any of Ally Lama’s replies, choose a reason and add a note if you like. The report includes the reply, the message it answered and your account ID, so we can find and review it (see our Privacy Policy).
• By email: write to support@allylama.com with the subject “Report a response”. If you can, include roughly when it happened and what was said.

A person reads every report, and we use them to improve Ally Lama’s safety. If a report is about someone’s immediate safety, please contact emergency services first.

How we keep this under review

The safety net is covered by automated tests that run before every change to it, and we test it against our live service. We review this protocol, the words it looks for and the helplines it shows regularly, and whenever we receive a report or the guidance changes. This page shows the date it was last reviewed.

For how we handle your data, see our Privacy Policy. For the rules of using Ally Lama, see our Terms of Service.