Safety and crisis protocol
Last updated: October 8, 2026
Aura is an AI, not a human. It can't provide therapy or medical care.
What is checked
Aura checks the words you type for signs of suicidal thoughts or self-harm: messages in AI chat (on the website, in the app and in the quick drawer), text messages you send Aura, and answers you give during voice onboarding. Journal entries are checked with the same patterns before anything in them is sent to AI. The check is a set of written patterns, not a person reading along and not a medical assessment.
Live voice
Live voice sessions are checked the same way: each thing you say is compared against the same patterns while the call is happening, and a match interrupts the call with the same crisis screen.
What you see
When the check finds a strong sign, Aura stops the normal reply and shows the crisis resources: “If you're thinking about suicide or self-harm, you can call or text 988 (988 Suicide & Crisis Lifeline) or text HOME to 741741 (Crisis Text Line). If you're in immediate danger, call 911.” You can call 988, text 988 or call 911 straight from that screen, and text HOME to 741741 for the Crisis Text Line. When the sign is milder (for example hopelessness with no mention of harming yourself), the reply goes ahead as normal with a resources card under it: tap to call or text the 988 Suicide & Crisis Lifeline, or to text HOME to 741741. In a text message the same resources are added to the reply; in a voice call Aura says them and the card appears on screen. Outside the US, contact your local emergency number.
How replies change
After a strong sign, the AI does not answer the message at all. You get a fixed safety message with the resources above instead of a generated reply, and the conversation does not continue on that topic until you send something else. Aura never gives instructions, means or methods related to self-harm.
What is recorded
Aura keeps an automated safety record that a check matched: the time and the severity. The words of your message are not stored in that record unless message storage is explicitly switched on for the service. Nobody is notified by it, and nobody monitors conversations in real time. Aura cannot contact emergency services for you.
Practitioners
If you are connected with a practitioner on Aura and a mood or wellbeing check shows a sustained pattern of distress, that can be flagged for the practitioner to review. That is separate from the crisis screen, it only goes to a practitioner you are already connected with, and it is not an emergency alert.
Reporting a harmful response
If an AI reply is harmful, wrong or distressing, email info@auragpt.ai with the subject “Report AI response” and the date and time of the reply. A person reviews every report. Reporting does not replace the crisis resources above.
Human review
Crisis checks themselves are automatic and are not reviewed by a person as they happen. Reports you send are reviewed by a person. The patterns, the wording of the crisis screen and this page are reviewed when they change.
Get help now
AI companions may not be suitable for some minors.