Safety and crisis protocol

Last updated: 22 September 2026

This page describes what Soulwise does when a conversation shows a risk of suicide or self-harm. We publish it so you can know in advance how the app behaves.

1. Soulwise is not an emergency service

Soulwise is an AI conversation companion. You are talking to an AI, not a human. It is not a crisis line, it does not provide health care, and it is not a substitute for a licensed professional. Nobody at Soulwise watches conversations in real time, and we cannot send help to you.

If you are in danger now, contact emergency services: 988 in the US and Canada (call or text); 112 in Türkiye and the European Union; in the UK, Samaritans on 116 123, or 999 in an emergency. Lines for other countries: Support.

2. How messages are checked

Every message you send for a reply is checked on our server before a reply is shown — the app cannot switch this off. Two checks run:

• OpenAI's moderation service, including its self-harm categories
• our own keyword checks for suicide and self-harm language

3. What happens when a risk is detected

• Instead of an ordinary AI reply, you receive a safety message with crisis resources for your region. This message is fixed text written in advance; it does not depend on what the AI would have said.
• Your region is taken from the country in your profile; if you have not set one, from the app language (English → US, Turkish → Türkiye). If your country is one we do not yet carry numbers for, we show no other country's number — a line that does not connect where you are is worse than none — and instead tell you to call your local emergency number and link to findahelpline.com, which lists lines for every country. Set your country in your profile so the right number is shown.
• In the app and on the website, crisis numbers are shown as tap-to-call links.
• For the next 72 hours the AI is instructed to be especially gentle in every conversation mode, to avoid confrontational framing, and to point you toward immediate help if distress shows again.

4. How the AI is instructed

In every conversation mode, the AI is instructed not to act as a clinician: it does not give professional health assessments, diagnoses or medication recommendations, and it points you toward a qualified professional if you ask for a diagnosis or a prescription.

Referral to crisis services does not rely on the AI alone: when a risk is detected, the safety message in section 3 is shown instead of the AI's reply. In addition, several conversation modes instruct the AI to step out of its usual style when someone speaks about suicide or self-harm and to point them to crisis services, and after a risk has been detected every mode is instructed to do so for 72 hours.

5. What is recorded

When a risk is detected, a safety record is written to your account that keeps only a category label (for example, "self-harm"), the conversation it came from and the time. The text of your message is not written to that record. The same label-only record is written when a message asks for a medical diagnosis or tries to override the AI's rules. Safety records are deleted automatically after 90 days, and immediately if you delete your account.

6. Reporting an AI reply

If an AI reply is harmful, inaccurate or inappropriate, report it:

• In the mobile app: press and hold the AI reply, then choose a reason.
• On the website, or if something else is wrong: write to soulwise@luminateknoloji.net.

When you report a reply, the text of that reply and any note you add are kept with the report for 90 days so we can review it.

7. Minors

Soulwise is for people aged 18 and older, and the age step at sign-up cannot be skipped: our server checks it before any conversation starts. AI companions may not be suitable for some minors. We close accounts we learn belong to someone under 18.

8. Contact

soulwise@luminateknoloji.net · Privacy Policy · Terms of Service · Support