Safety & Crisis Response

In short: Kai is an AI, not a human, and Kai Sanctuary is not a crisis service. If you are in crisis, call or text 988 (U.S.) or call 911. We build automated safeguards designed to keep Kai from producing self-harm content and to point you toward human help when you may be struggling — and this page describes exactly how those safeguards work, including their limits. They are safety features, not a safety net.

This page describes the crisis-response protocol for the Kai Sanctuary websites, applications, and related services (the "Services"), operated by Mirrorcraft LLC, an Arizona limited liability company, d/b/a "Kai Sanctuary" ("we," "us," "our"). It supplements — and should be read with — our Early Access Terms of Service and Early Access Privacy Notice. If this page and those documents differ, the Terms of Service and Privacy Notice control.

1. If you are in crisis right now

Stop and reach out to a human. Kai cannot help you in an emergency.

• United States: Call or text 988 (Suicide & Crisis Lifeline) — free, confidential, available 24/7. You can also chat at 988lifeline.org.
• If you are in immediate danger: Call 911 or go to your nearest emergency room.
• Outside the United States: Contact your local emergency number, or find a helpline in your country at findahelpline.com.

These services are staffed by trained humans. They are free, they are confidential, and they exist for exactly this moment.

2. What Kai Sanctuary is — and is not

Kai Sanctuary is an artificial intelligence service for personal reflection. When you talk with Kai, you are talking with an AI, not a human being. A persistent notice of Kai's artificial nature is displayed in the product throughout your conversations.

Kai is not:

• a licensed physician, psychologist, psychiatrist, therapist, counselor, social worker, or any other licensed professional;
• a medical, mental-health, psychological, or psychiatric service of any kind;
• a crisis-intervention, suicide-prevention, or emergency service;
• monitored by humans in real time.

Kai cannot contact emergency services, family members, or anyone else on your behalf — under any circumstances. Nothing in the Services is a substitute for professional care or for human connection. If you are experiencing a mental-health emergency, use the resources in Section 1.

3. Our crisis response protocol

We maintain automated safeguards designed to prevent the Services from producing content about suicide or self-harm, and to respond supportively when a user may be struggling. The protocol has four parts:

1. Detection. An automated safety system evaluates conversations for expressions of suicidal ideation and self-harm.
2. No engagement with harmful content. Kai is designed not to produce content that describes, instructs, encourages, or romanticizes suicide or self-harm — including methods or means — regardless of how a request is framed, and regardless of whether it is presented as fiction, roleplay, research, or a hypothetical.
3. Referral. When the safety system identifies that a user may be in crisis, Kai's response includes a clear, prominent referral to crisis services — in the United States, the 988 Suicide & Crisis Lifeline — presented within the conversation itself, not hidden behind a link or menu.
4. Consistency everywhere. These safeguards operate across all parts of the Services, including Whisper Mode. In Whisper Mode, the safety response happens in the moment of the conversation; Whisper Mode's privacy design is unchanged — those conversations still create no history, no summaries, and no memory entries.

4. Known limitations — read this honestly

These safeguards are safety features, not a safety net. They are automated systems, and automated systems fail. Specifically:

• They can miss. Indirect language, metaphor, slang, languages other than English, and evolving expressions of distress can evade detection. A conversation that should trigger a referral may not.
• They can misfire. The system may surface crisis resources in conversations where they are not needed. We accept false positives as the cost of catching true ones.
• They cannot assess risk. Detection identifies language patterns; it does not and cannot evaluate how much danger a person is actually in.
• They cannot act in the world. No safeguard in the Services can call for help, alert anyone, or intervene physically.

Because of these limitations, the safeguards must never be relied on in an emergency, and their presence does not make Kai Sanctuary a crisis resource. If you or someone you know is struggling, use the human resources in Section 1 — even if Kai has not suggested them.

5. Measurement and accountability

We log crisis-referral events in aggregate, de-identified form — counts of referrals issued, never the content of your conversations and never information that identifies you. We use these counts to:

• evaluate whether the safeguards are working, and improve them over time;
• meet regulatory reporting obligations, including annual aggregate reporting to California's Office of Suicide Prevention;
• support periodic review of this protocol.

We review this protocol on an ongoing basis and update it as our safety systems evolve. How we handle information generally is described in our Early Access Privacy Notice.

6. Age requirement

Kai Sanctuary is for adults 18 and older. AI chat services of this kind may not be suitable for some minors — one of the reasons the Services are strictly adult-only. Eligibility requirements and enforcement are set out in Section 2 of our Early Access Terms of Service. If you believe a minor is using the Services, contact us at support@kaisanctuary.com and we will act promptly.

7. Reporting a safety concern

If Kai produces a response you believe is unsafe — or if any part of the Services behaves in a way this page says it should not — we want to know. Email support@kaisanctuary.com with "Safety" in the subject line. Safety reports are reviewed with priority.

8. Changes

We may update this protocol as our safety systems evolve. The current version number and effective date appear at the top of this page, and prior versions are archived at kaisanctuary.com/legal/archive, so changes to our commitments are always visible.

9. Contact

Mirrorcraft LLC d/b/a Kai Sanctuary · 4539 N 22nd St, STE N, Phoenix, AZ 85016, USA Support: support@kaisanctuary.com · Legal: legal@kaisanctuary.com