Crisis Protocol

Vita is an AI health information app. It is not a person, a doctor, a therapist or a crisis service. This page says what Vita does when a message suggests a medical emergency or self-harm, and what it does not do.

Every question you send, typed or spoken, is checked before an answer is written. A spoken question is checked from its transcript.

• A fixed list. Our server first looks for clear self-harm phrases, from a fixed list in many languages. No AI is involved in this step, and it runs even if the AI service is down.
• A small AI check. If the list finds nothing, a small call to our AI provider (Privacy Policy, section 9) reads the message and labels it. Two labels matter here: an emergency happening now to you or someone with you, and thoughts of self-harm or of harming someone else. Short greetings skip this step.
• Limits never come first. The free daily limit and the voice limits do not stand in front of these checks. A message that reads as an emergency or a crisis goes through even when the limit has been reached.

If it reads as an emergency, Vita shows a fixed message instead of an answer. It tells you to call your local emergency number now, gives the number, and adds one short first-aid tip for while you wait.

If it reads as self-harm, Vita shows a fixed message instead of an answer. It says that what you are feeling is real and that trained people can help, and it asks you to call. It gives the crisis line we hold for your country. Where we hold none, it gives the emergency number.

No AI writes either message. They are written in advance, in the languages of the app, and shown in the language of your message where we can tell it. In voice mode, Vita reads the message aloud.

The number comes from the country your connection is in at that moment, so someone travelling gets the local number. If that is unknown, we use the country on your account. If we hold no number for a country, Vita shows 112, which is not the emergency number everywhere.

• It does not use up a free answer.
• The app does not put a plan offer or a request to rate Vita under that message.
• In the iPhone app, version 3.6.3 and later, Vita suggests no next step for the following hour, and nothing in that message is saved to a profile.
• Your next message is checked in the same way. If it does not read as an emergency or a crisis, Vita answers it normally, and the instructions for that answer carry your country’s crisis line and emergency number.

• It does not call anyone. Vita does not contact emergency services, a crisis line, your family, your doctor or anyone else.
• No one is alerted. No one at VitaMedas is told when these messages appear, and no one watches conversations as they happen.
• The checks can be wrong. They are automatic. They can miss a message that needed help, and they can show an emergency message for one that did not.
• The list covers self-harm only. Emergencies depend on the AI check. If that check fails, the message goes on to an ordinary answer.
• The checks read your words. They do not read the inside of a file you attach.
• It is not crisis support. Vita does not replace a crisis line, a doctor or emergency care.

Your message and the reply are saved in the conversation like any other, with a mark showing which kind of reply it was. That mark is one of the safety signals in the Privacy Policy. You can delete the conversation. We use counts of these replies to check that the system works. Who at VitaMedas can read a conversation, when, and how each read is logged, is in section 7.4 of that policy.

We test these checks with sample messages in the languages Vita supports. A daily automated test also runs sample messages through the AI check, to look for missed emergencies and for false alarms.

We update this page when these checks change. Questions: support@vitamedas.com. Please do not use email in an emergency.