Safety protocol

Effective 13 October 2026. This page describes what Busyline does when a conversation suggests someone may be thinking about suicide or self-harm. It applies to the app as designed for launch; Busyline is not released yet.

How the app notices

• A pattern screen. What you just wrote or said is checked against a list of clear phrases in the eight languages the app supports (English, Spanish, French, German, Arabic, Japanese, Portuguese and Hindi). The check runs on our server, in memory, and the text is not kept.
• The AI model's own judgment. Every conversation carries a safety instruction that overrides the character. The model tells the app when a message suggests suicide, self-harm or immediate danger, or deep distress or hopelessness, and it tells ordinary frustration, jokes and exaggeration ("this date is killing me") apart from the real thing.
• The two are combined and the more cautious result wins. A clear statement about ending one's life is always treated as high concern.

What happens when it does

• The reply changes. The character stops joking, flirting and role-playing and answers warmly and briefly in your language. It acknowledges how you feel, says plainly that it is an AI and cannot replace a person, encourages you to reach out to someone you trust or a crisis line, and asks whether you are safe right now. It does not describe methods, argue, promise secrecy, change the subject or leave. It is never sent as a voice note.
• The app shows crisis resources right under the reply: a crisis line for the country your phone is set to, with call and text buttons where the line offers them, a button for the emergency number you chose in the app, and a link to a directory of helplines in other countries. Every button only opens your phone's dialer, messages app or browser. Nothing dials or sends anything by itself.
• If the AI refuses or fails to answer a message that was flagged, a fixed, pre-written supportive message is used. It never falls back to a casual deflection.
• On a voice call, the call stays open. It is never wrapped up or ended by the character, the "reason to leave" plan is dropped, and the same resources stay on screen.
• Help is always one tap away. The chat screen has a "Get help now" button that works at any time, whether or not anything was flagged, and does not need an internet connection.

Reminding you it is an AI

Every chat shows a notice at the top that the character is an AI. During a call, the screen says "AI character" the whole time. And while you keep chatting, the app adds a dated reminder to the conversation every three hours ("Reminder: Sam is an AI, not a person."), in your language, so it cannot scroll out of view.

What Busyline does not do

• It does not contact anyone on your behalf, share your location or alert emergency services, friends or family. It has no way to.
• It does not store flagged messages or keep a record of you. The server records only the level of concern and whether it came from chat or a call, never the text or who said it. Chats stay on your device.
• It does not diagnose anything and is not therapy.
• It can miss things and it can raise a flag when you were only venting. Neither makes it a safe replacement for a person or a crisis line.

Crisis lines the app offers

These are the numbers built into the app, checked on 13 October 2026. Hours and availability can change. If your country is not listed, the app links to the Find A Helpline directory, and your own emergency number is always shown.

Reviewing this protocol

We review the screening phrases, the AI instruction, the wording and the list of crisis lines at least every six months, and whenever a service changes. If you spot a wrong number or a way the app handled a hard moment badly, please write to hello@busyline.app.

Who this is for

Busyline is for adults, 18 and over. If you are under 18 and need help, please talk to a trusted adult or call your local youth or emergency line.