Safety & Crisis Protocol

1. What ReasonCare is

ReasonCare's companion is an AI. It is not a person, not a therapist, doctor, or medical or mental-health service, and it does not diagnose or treat any condition. We say so at signup, at the start of every conversation, in the app's settings, and again every three hours during a long conversation. Our clinic product is different: there, a licensed clinician at your clinic directs the care and reviews what is shared.

2. How we watch for crisis

Every message a person sends is screened, on our servers, for signs of suicidal thinking, self-harm, or danger to others, using layered detection: a fast keyword and phrase screen, a dedicated safety classifier, and the conversation model's own judgment. The companion is not permitted to produce content that encourages or instructs self-harm or suicide, and it is instructed to move toward real-world help whenever risk appears.

3. What happens when risk is detected

• The companion responds with care, stays with the person, and refers them to crisis services. A crisis card appears in the app with 988 and other resources, and a one-tap option to call.
• Where the person has agreed to it in advance, the app can place the call to 988 directly.
• If the person stops responding after a high-risk signal, the app re-surfaces the crisis card and, if they added an emergency contact and consented, that contact can be notified.
• For clinic patients, the event is recorded and flagged to the care team through the clinic's own system so a clinician can follow up.

4. Human review

Crisis events and every user report of an AI response are reviewed by people on our team. Anyone can flag a response from inside a conversation (long-press on the phone, “Report” on the web). We use what we learn to correct the companion's behaviour, and we do not use it for advertising.

5. Minors

The companion is for adults 18 and older. We ask for year of birth at signup, use age information provided by the app stores where it is available, and close accounts that misrepresent age. People under 18 use ReasonCare only through a participating clinic, under the direction of a licensed clinician and with any consent the law requires.

6. Limits

Automated detection is not perfect. It can miss risk and it can raise it when none exists. ReasonCare does not monitor conversations in real time with staff, and it cannot dispatch emergency services. If you or someone you know is in immediate danger, contact emergency services.

7. Reporting and contact

To report a safety concern about ReasonCare, email safety@reasoncare.ai. See also our Terms of Service and Privacy Policy.