Last updated: [EFFECTIVE DATE]
DRAFT. To be reviewed by a licensed clinician and counsel before launch. This page describes exactly what Upbee does today. Update it whenever that changes.
The short version
- Upbee is an AI, not a person. It says so whenever you ask, and it never pretends otherwise.
- Upbee is not an emergency service. No human watches conversations in real time, and Upbee can't call anyone for you.
- If you're in danger right now, call 911. If you're thinking about suicide or self-harm, or you're worried about someone else, call or text 988 (the Suicide & Crisis Lifeline) or text HOME to 741741 (Crisis Text Line). Both are free and open 24/7.
1. Who Upbee is for
Upbee is for adults 18 and older. We confirm age at signup and don't knowingly serve anyone younger.
2. What happens if you text something that suggests a crisis
Every message you send is checked before it reaches the AI:
- Detection. Upbee looks for language suggesting a risk of suicide or self-harm, for example talk of wanting to die, hurting yourself, or not wanting to be alive.
- A fixed, human-written response. When that language appears, the AI does not improvise a reply. Upbee sends a short message written in advance by people. It says we're glad you reached out, gives you 988 (call or text), Crisis Text Line (text HOME to 741741) and 911 for immediate danger, and lets you know you can keep texting.
- No pressure, no lecture. Upbee doesn't diagnose, argue, or try to act as a counselor. It points you to people who can help, right away.
- You stay in control. You can keep using Upbee normally afterward. Nothing gets locked or reported to anyone you know.
We tune this to catch real risk even if it occasionally responds to a message that wasn't one. Missing someone who needs help is the worse mistake.
3. What we record
- When the crisis response is sent, we log only that it happened and its category, never the content of your message, so we can count how often it's used and improve it.
- We don't share crisis events with advertisers (we have none) or with anyone you know.
- [If required by law: aggregate, de-identified counts of crisis-referral notifications are reported annually to the California Office of Suicide Prevention, beginning July 1, 2027. They contain no personal information.]
4. How the AI is instructed
Separately from the check above, the AI's own instructions tell it to:
- respond with care and keep it short if a conversation turns to self-harm or danger, and point to 988 or 911;
- never encourage self-harm, harmful behavior or isolation;
- never claim to be human or a licensed professional;
- give general information on health, legal or money topics and encourage a professional for decisions.
5. Limits you should know about
- Detection isn't perfect. It can miss indirect language, and it can't understand images or voice notes the way a person would.
- Messages can be delayed by carriers or networks. Never rely on Upbee in an emergency. Call 911 or 988 directly.
- Upbee doesn't know where you are and can't send help to you.
6. Getting help for someone else
If you're worried about someone, you can call or text 988 for guidance on how to help, any time.
7. Contact
Questions about this protocol: [safety@upbee.ai]. This page is reviewed at least once a year and whenever the protocol changes.