Tenari includes an AI Companion that holds an ongoing personal conversation. We publish this protocol so anyone can see exactly what happens when a user expresses suicidal thoughts or self-harm, as required of companion-chatbot operators under California SB 243 and New York General Business Law Article 47.
Every message a user sends to the Companion is checked, before the AI answers, against a published set of language patterns covering four categories modelled on the risk levels used in the Columbia-Suicide Severity Rating Scale:
This check runs on our own servers and does not depend on the AI model or on any outside provider, so it keeps working during a provider outage. It deliberately errs toward offering help: an ambiguous message is more likely to produce a referral than to be passed over. It suppresses matches where the user is clearly describing an accidental physical injury, or discussing another person rather than themselves.
When a referral is shown we store three things: the category, the timestamp, and which app surface it came from. We do not store who the user was, and we do not store the message that triggered it. There is no per-user record of suicidality anywhere in our systems, by design — such a record would be one of the most sensitive datasets we could hold, and we do not want it.
The trade-off is deliberate and we state it plainly: because the record carries no identifier, we cannot follow up with an individual user, and we cannot tell you whether a referral was ever shown to you.
The count of referrals is what we report annually to the California Office of Suicide Prevention beginning 1 July 2027, alongside a description of these protocols. That report contains no user identifiers and no personal information, because we do not collect any against these records.
Tenari tells you that you are talking to an AI when a conversation begins and again at least every three hours of continued use. The AI is never presented as a human, a therapist, a counsellor, or a clinician, and it is not permitted to claim to be one.
Tenari is for adults aged 18 or over. We collect a date of birth at signup and refuse registration below 18. A companion chatbot may not be suitable for some minors. If we learn that an account belongs to someone under 18, we remove it — see Privacy Policy §10.
If the Companion ever responded unsafely to someone in crisis, tell us and we will treat it as the highest-priority defect there is: containedevolution@gmail.com.
This protocol is a working beta document and is subject to review by qualified counsel before general availability. It will be updated as the safeguards change; the version at the top of this page identifies the one in force. See also the Terms & Conditions and Privacy Policy.