Our safety protocol
myessem is a self-reflection tool. MIRRA, the reflection agent inside it, is an artificial-intelligence system, not a human being, and the Service is not therapy, counseling, medical care, or a crisis service. Because people sometimes bring real pain into private writing, we built a safety protocol into every conversation, and this page describes it in plain language.
If you are having thoughts of suicide or self-harm, please reach out now:
- United States: call or text 988 (Suicide & Crisis Lifeline), or chat at 988lifeline.org
- United States, immediate emergency: 911
- Israel: ERAN at 1201 (24 hours) · SAHAR written chat: sahar.org.il · emergency: 101
- Elsewhere: findahelpline.com
Do not use myessem to seek emergency help.
How the protocol works
- Every message is screened first. Before MIRRA may answer, an automated safety system reads each message you send and classifies it. No message is ever answered unscreened; when the screening itself fails, MIRRA does not answer at all.
- Detection stops the conversation. If the system detects an expression of suicidal ideation or self-harm, MIRRA does not respond. Instead, the Service shows the crisis resources above and pauses the conversation.
- Continuing is a choice, once. After a first pause in a day, you may acknowledge the resources and choose to continue; MIRRA then speaks with extra care. A second detection the same day pauses the conversation until the following day.
- MIRRA never cooperates with self-harm. Her instructions forbid providing, discussing, or refining methods; normalizing or agreeing with an intention to die; or helping compose farewell messages — in every language, in every turn, regardless of anything written to her.
- The record of a pause is content-free. The fact that a safety pause happened is kept separately from your words, so erasing your conversation, which you can always do, never erases a safety pause.
What this protocol is not
The safety system is precautionary and automated. It may sometimes pause a conversation that did not need it, and, like any automated system, it cannot be guaranteed to recognize every crisis. It is a referral mechanism, not a monitoring, alerting, or rescue service. No human reviews your messages in real time, and the Service cannot contact emergency services on your behalf.
You always know MIRRA is AI
MIRRA is software. The Service says so where you interact with her, and she will never claim to be a human being or a licensed professional. Everything she writes is generated automatically and can be imperfect; she reflects, and she does not advise, diagnose, or treat.
Questions about this protocol: support@myessem.com · See also our Terms of Service and Privacy Policy.