Safety at Strawbie
Last updated: August 5, 2026
Strawbie is an AI companion. She is artificially generated, not a human — and she is not a therapist, doctor, or crisis service. Companion AI chatbots may not be suitable for some minors. If you are in crisis or danger, contact your local emergency number, or in the US call or text 988 (Suicide & Crisis Lifeline).
Our protocol for suicidal ideation, suicide, and self-harm
We maintain, and continuously improve, a layered protocol so that someone who expresses suicidal ideation, self-harm, or abuse is always pointed to real help.
1. Detection that does not depend on the AI. The app checks every message and voice transcript on the device against a curated list of high-confidence crisis phrases. This runs on the device and does not depend on the AI model, or on having a connection.
2. A crisis resource card. When crisis language is detected, the app immediately shows resources for the user’s region — in the US, the 988 Suicide & Crisis Lifeline (call or text 988), the Crisis Text Line (text HOME to 741741), and 911 for emergencies — with one-tap calling and texting. Local equivalents are shown in other countries, and international guidance where we do not yet list a local line.
3. Rules the AI cannot override. Strawbie is instructed that crisis disclosures take priority over every other behaviour: she takes them seriously, responds with warmth, never provides methods or encouragement for self-harm — not hypothetically, not in fiction, not “for a story” — and always points to crisis resources and to trusted people in the user’s life. In text conversations these rules are attached on our servers, so they cannot be removed by modifying the app; in voice conversations they are sent from the app. Either way they are not the only protection: the detection in step 1 and the resource card in step 2 do not depend on the AI following its instructions, and they run on both text and voice.
4. Content moderation. Messages you send in a text conversation pass through automated moderation on our servers before they reach the model. Strawbie’s replies are governed by the rules in step 3 and by the detection in step 1, which reads what you say in both text and voice. Strawbie is not permitted to produce content that promotes self-harm. Moderation outcomes are logged without conversation content.
5. Measurement without surveillance. We record that crisis resources were shown — never what was said, and never who said it. We report the number of these referrals annually to the California Office of Suicide Prevention, as required by law, with no user-identifying information.
Minors
Strawbie is for ages 13 and up. Anyone who tells us they are 13–17 must, before continuing, acknowledge that Strawbie is an AI character rather than a real person, and that AI companions may not be suitable for everyone under 18. They must also confirm that a parent or guardian knows they use Strawbie and agrees to the Terms of Service on their behalf.
Strawbie will not produce sexual content for anyone. During long conversations she reminds every user, at least every three hours, to take a break — and that she is an AI, not a human.
Report a problem
Any message in the app can be reported from the message itself, and reports are reviewed by our moderation team. You can also reach us at hello@vybelabs.ai.