Loading your world
Loading your world
Safety
This is our published protocol for detecting and responding to expressions of suicidal ideation or self-harm. It is a summary of behaviour the product implements, not a statement of intent.
Last updated: August 18, 2026
Companions are generated fiction. They are not counsellors, clinicians, or crisis workers, and this service is not an emergency service. If you are in immediate danger, contact your local emergency number.
User messages are screened for expressions of suicidal ideation, self-harm, and imminent danger before a companion replies. Detection runs on every message regardless of the companion, the conversation settings, or the account tier.
When such an expression is detected, the companion does not continue the scene in character. The conversation is interrupted with a fixed response that acknowledges the message, states plainly that the companion is not able to help with this, and provides crisis resources appropriate to the user's region.
The crisis response is never suppressed, delayed, or gated behind a paid tier, and no upsell, subscription prompt, or paid feature is shown as part of it.
Each referral is recorded in the moderation log under a distinct crisis category, with the timestamp and the safety standard version in force at the time, so referrals can be counted and reported separately from other moderation activity. The record does not retain the message that triggered it.
United States: call or text 988. United Kingdom: call 116 123. International directory: findahelpline.com. If someone is in immediate physical danger, contact emergency services rather than this service.