How the hub responds when someone is in crisis

1. The answers come from A.I., not a person

On the hub's page, people in California see a notice that each answer is produced by the A.I. model of the provider they choose, not by a person.

2. When someone expresses thoughts of suicide or self-harm

The hub reads each message for words that express thoughts of suicide or self-harm, in English and Spanish. When it finds them, the answer comes with a referral to crisis services, shown first:

• the 988 Suicide and Crisis Lifeline: call or text 988, or chat at chat.988lifeline.org, free, confidential and open 24 hours a day (text and chat are also offered in Spanish);
• 911, for anyone in immediate danger.

The referral is shown even when the message cannot be sent, for example when the wallet is empty.

3. Keeping harmful content out of answers

Every model that answers a person in California is told never to encourage suicide or self-harm, never to describe methods of it and never to give instructions for it, and to point a person in crisis to 988.

When a message expresses thoughts of suicide or self-harm, the model's answer is read by Govern All A.I.'s own checking model before the person sees it. An answer that encourages suicide or self-harm, describes a method or gives instructions is held back and is not charged, and so is any answer the check cannot read.

4. Minors

Companion chatbots may not be suitable for some minors. The hub is for adults: everyone confirms they are 18 or older before using it.

5. What we count and report

Each referral is counted with its date and place only, never the person or the message. Each year from July 1, 2027, Govern All A.I. reports to California's Office of Suicide Prevention the number of referrals given in the year before and these protocols, as section 22603 asks, with no identifier or personal information.

6. Questions

Write to [email protected]. If you or someone you know is in crisis, call or text 988 now.