Suicide and self-harm: how Margo responds

If you need help now

What Margo does

1. It notices. Before Margo answers, what you just wrote or said is checked against a fixed list of phrases about suicide, self-harm and not wanting to live, in English and German. This runs on every message, on the iPhone, the Mac and in phone calls.
2. It refers you to people who can help. If the check matches, Margo’s answer starts with crisis lines in your language — before anything else — and says plainly that it is an AI and cannot help in an emergency.
3. It stays careful. Margo’s AI is instructed by its provider not to encourage or describe self-harm; Margo does not give instructions or methods, and it does not try to keep you talking instead of reaching a person.
4. It tells your trusted contact — only if you chose one. If what you write states an intention or a plan to end your life, and you have named a trusted contact who has accepted, Margo sends that person one email saying you may need support and how to help. It never includes anything you wrote. At most one such email a day. You can add or remove a trusted contact in Margo’s settings; they must accept before anything is ever sent to them.

What is recorded

Only a count: how many times, per day, Margo showed crisis lines and at which level. No account, no words, no time of day. California requires a yearly report of these numbers from July 2027. Your conversation itself is kept like any other conversation in Margo, and you can delete it.

Limits

• The check works on words. It can miss someone in danger who says it differently, and it can show crisis lines to someone who is fine. We would rather do the second.
• Margo does not call emergency services and cannot see where you are.
• Margo reminds you that it is an AI: at the start, by the badge beside its name, and every three hours in a long conversation.
• Margo is only for adults aged 18 and over.

Questions

Write to help@margo.global. How Margo uses AI · Privacy notice · Consumer health data privacy policy.